Slashdot Mirror


SoHo NAS With Good Network Throughput?

An anonymous reader writes "I work at a small business where we need to move around large datasets regularly (move onto test machine, test, move onto NAS for storage, move back to test machine, lather-rinse-repeat). The network is mostly OS X and Linux with one Windows machine (for compatibility testing). The size of our datasets is typically in the multiple GB, so network speed is as important as storage size. I'm looking for a preferably off-the shelf solution that can handle a significant portion of a GigE; maxing out at 6MB is useless. I've been looking at SoHo NAS's that support RAID such as Drobo, NetGear (formerly Infrant), and BuffaloTech (who unfortunately doesn't even list whether they support OS X). They all claim they come with a GigE interface, but what sort of network throughput can they really sustain? Most of the numbers I can find on the websites only talk about drive throughput, not network, so I'm hoping some of you with real-world experience can shed some light here."

517 comments

  1. SMB by Anthony_Cargile · · Score: 0

    The network is mostly OS X and Linux with one Windows machine (for compatibility testing)

    Well it looks like SMB is your best bet for compatibility. For a budget, just go with a small Linksys or Cisco device, as you can specify the hard drive and the network around it governs the speed.

    1. Re:SMB by Anthony_Cargile · · Score: 4, Informative

      One more thing: if it says gigabit ethernet, for me that usually means anywhere between 200-800Mbps of speed on a fairly busy network, which should suffice for large data backups in a matter of say 2-5 minutes tops for moving several gigs. Your throughput really depends on other factors, so yours may be higher or lower than mine but typically that range should suffice with the proper switching and routing equipment.

    2. Re:SMB by drsmithy · · Score: 1

      Well it looks like SMB is your best bet for compatibility. For a budget, just go with a small Linksys or Cisco device, as you can specify the hard drive and the network around it governs the speed.

      This isn't really true. For *lots* of low-end NAS devices, the performance limitation is their puny CPUs, that can barely shift bits fast enough to saturate a 100M link.

    3. Re:SMB by SanityInAnarchy · · Score: 2, Insightful

      Well it looks like SMB is your best bet for compatibility.

      OS X doesn't support NFS? Linux doesn't support AFP?

      Besides which, don't the better NAS boxes support pretty much everything, all at once?

      --
      Don't thank God, thank a doctor!
    4. Re:SMB by Anthony_Cargile · · Score: 1

      Depending on their budget, linksys/netgear is cheaper, yet trades off quality as you mentioned due to lower-end hardware. For a little more (going a long ways) a smaller Cisco NAS would suffice as an out-of-the-box solution that does not sacrifice speed and throughput (depending on the model, of course).

      A custom-built box, as many commenters suggested, seemed a tad inappropriate to me as he asked for an NAS device, not a server. Installing Ubuntu or whatever on it seems like more of a performance hit than a properly optimized "off the shelf" NAS box, since they most likely don't run Dbus, GNOME, Hald, bluetooth or any other desktop software atop the basic kernel and networking services.

    5. Re:SMB by drsmithy · · Score: 4, Informative

      A custom-built box, as many commenters suggested, seemed a tad inappropriate to me as he asked for an NAS device, not a server. Installing Ubuntu or whatever on it seems like more of a performance hit than a properly optimized "off the shelf" NAS box, since they most likely don't run Dbus, GNOME, Hald, bluetooth or any other desktop software atop the basic kernel and networking services.

      While this is true, for noticably less than you'll pay for a NAS appliance, you can build a PC with vastly more CPU power and RAM (in particular, storage vendors - even with high-end, full-blown SAN solutions - are offensively stingy with cache), which will more than make up for any extra stuff that might be running.

      You need to spend a LOT on an "appliance" type storage system to get something that has higher performance and/or better features than a "server". Particularly with cache, storage vendors across the board are offensively stingy (16 gigs of high-quality ECC RAM costs maybe $800, but you'll be lucky if your $100k SAN comes with half that amount).

      Personally I would recommend the OP looks at Server/NAS-style "appliances" like Dell's NF500. They're the only sort of "cheap" turnkey devices he'll find that will deliver the performance he seems to want, and will probably only cost a grand or two more than DIY.

    6. Re:SMB by drsmithy · · Score: 1

      OS X doesn't support NFS? Linux doesn't support AFP?

      Not as well as they (respectively) support SMB.

    7. Re:SMB by LoRdTAW · · Score: 4, Informative

      A NAS is pretty much a server that is dedicated to storage.

      If he wants to roll his own I would suggest either a light install of Ubuntu server or FreeNAS: http://www.freenas.org/. FreeNAS is based on the stripped down Free BSD core that m0n0wall uses. It is very small and is managed using a simple and easy to use web interface. I don't know about gigabit performance as I only set it up once for a friend using 100mbit. He had the Linksys NAS box and it was dog slow. On 100Mb it couldn't push more then 3-4 MB sec. I could get 8-9Mb sec using FreeNAS on an Athlon 1.3Ghz with 128MB ram and two SATA 500GB drives in RAID 1 (mirroring). He also added a USB 2.0 card to hook up another 500GB drive. It pretty much saturates his 100Mbit connection.

      And here is my related question to others here:
      I have fought with SAMBA on Ubuntu 8.04 server and I cant get it going faster than 10-11MB/sec when copying to/from Windows XP. Even with the tcp_nodelay setting and a few others it just barely breaks 11MB/sec. I can get 25-30MB sec when copying from one Windows PC to another. And the server hardware isn't puny: dual P4 2.4GHz Xeons, 4GB RAM, dual PCIX Intel gigabit and a PCIX SATA controller. Any one have any suggestions? NFS also runs at the same speed and when downloading from the Apache server I get 5-6MB sec. Something is wrong somewhere but I cant tell. I have changed kernels played with conf files but nothing works. Someone once told me SAMBA will always be slow but I don't believe that to be true.

    8. Re:SMB by iamhassi · · Score: 1

      "A custom-built box, as many commenters suggested, seemed a tad inappropriate to me as he asked for an NAS device, not a server."

      While that might be true in certain circumstances, the article does say "I work at a small business where we need to move around large datasets regularly...network speed is as important as storage size."

      This is for a business, not a home fileserver to share pictures and videos of the family vacation. If network speed really is a top priority then nothing will beat a custom-built box to be used as a server.

      --
      my karma will be here long after I'm gone
    9. Re:SMB by Anonymous Coward · · Score: 2, Insightful

      Saying "Gigabit ethernet" means nothing. For instance, Intel SS4000-E comes with "dual gigabit ethernet ports". Wow. This must mean that it supports up to 2Gbps, right?

      Wrong.

      First, the two ports don't support link aggregation, they're independent. Second, instead of a real-world performance of about 50-70 MB/sec on a gigabit link, this unit gives you... wait for it... 5 to 10 MB/sec.

      That's right, no typo there. Its CPU is so sleezy that that's all it can manage on small files. Large files get you up to 15MB/sec.

    10. Re:SMB by Anonymous Coward · · Score: 0

      I haven't used Leopard in production yet, so maybe things have changed... However, with Tiger I have had poor speeds when connecting OSX to SMB mounts.

      Sorry, I wish I had numbers to back it up. Also, it's possible I'm confusing SMB and FAT32 access (maybe I'm getting old... :'( )

    11. Re:SMB by bkeeler · · Score: 2, Informative

      run "ethtool eth0" and have a look at the output. It's possible that it's autonegotiated a stupid setting like half-duplex or some lower speed.

      Do the same with the windows box; that information is the properties dialog for the network device.

    12. Re:SMB by Anonymous Coward · · Score: 1, Informative

      A custom-built box, as many commenters suggested, seemed a tad inappropriate to me as he asked for an NAS device, not a server. Installing Ubuntu or whatever on it seems like more of a performance hit than a properly optimized "off the shelf" NAS box, since they most likely don't run Dbus, GNOME, Hald, bluetooth or any other desktop software atop the basic kernel and networking services.

      Most "Properly optimized" NAS units run a cut back Linux system anyway. There's no magic in them.

      You can also install Ubuntu Server (or any other distro for that matter) without Dbus, GNOME, Hald, bluetooth, etc. I know it's an amazing concept for some to be able to install an OS in a custom way, rather than vendor-enforced configs. That would be the "free as in freedom" part these Linux kids keep rabbiting on about.

    13. Re:SMB by Bearhouse · · Score: 2, Informative

      I have fought with SAMBA on Ubuntu 8.04 server and I cant get it going faster than 10-11MB/sec when copying to/from Windows XP. ...Someone once told me SAMBA will always be slow but I don't believe that to be true.

      Well, for SAMBA tuning, try (pdf):

      http://tinyurl.com/5rfjvu

      Alternatively, if you don't need all the Win network support that SAMBA provides, you can install ext2ifs on the XP boxes and enjoy easy and fast access to your *nix volumes. Works well for me. Caution: Security issues...

      http://www.fs-driver.org/index.html

    14. Re:SMB by Fulcrum+of+Evil · · Score: 1

      yeah, go get a 2u server with a 3ware raid card - $2000 and it can run at disk speeds.

      --
      "We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
    15. Re:SMB by Anonymous Coward · · Score: 1, Informative

      Depending on their budget, linksys/netgear is cheaper, yet trades off quality as you mentioned due to lower-end hardware. For a little more (going a long ways) a smaller Cisco NAS would suffice as an out-of-the-box solution that does not sacrifice speed and throughput (depending on the model, of course).

      A custom-built box, as many commenters suggested, seemed a tad inappropriate to me as he asked for an NAS device, not a server. Installing Ubuntu or whatever on it seems like more of a performance hit than a properly optimized "off the shelf" NAS box, since they most likely don't run Dbus, GNOME, Hald, bluetooth or any other desktop software atop the basic kernel and networking services.

      Thats why you when run a server, you don't install GNOME or anything like that on it...

    16. Re:SMB by Anonymous Coward · · Score: 0

      Make sure you are using something besides the generic pci ide drivers.

    17. Re:SMB by ingo23 · · Score: 1

      I got some noticeable (but still not reaching Win-to-Win speeds) improvements on Gigabit network when I replaced the NIC (and kernel modules, obviously). For some reason motherboard built-in Gigabit adapters are getting better results. I also noticed some strange degradation of HD performance when LVM is built on top of RAID (all Ubuntu server 8.10).

    18. Re:SMB by Kent+Recal · · Score: 1

      Amen for FreeNas.
      A NAS appliance that can saturate a GBit link (or two) is usually advertised as such. You don't find these things in the low-end because Joe Sixpack doesn't care and the components to support GigE ain't cheap by "cheap NAS standards".

      Thus your choices are to buy a "real" NAS from a real vendor that you can talk to, a low-range device will run in the $1500 ballpark, without drives.
      Or you just get/assemble an off-the-shelf beige box, add drives, throw FreeNas on it, configure half a day and be done. This should be doable in the $600 range, depending on your capacity requirements.

    19. Re:SMB by pyite · · Score: 1

      yeah, go get a 2u server with a 3ware raid card - $2000 and it can run at disk speeds.

      Or just use commodity SATA and ZFS. Much better than 3ware.

      --

      "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

    20. Re:SMB by MarcQuadra · · Score: 2, Informative

      8-9 MB/Sec? Really?

      I was getting 45-60MB/Sec (basically drive speed) on an old dual-cpu 1Ghz Pentium 3. I had Linux and Samba and no GUI running on it.

      Try throwing a low-end dual Core 2 (like an E5200) in an Intel board with a recent ICH chipset. Choose some -quality- drives, like WD RE3s, and a good network switch, like an SMC 8508-T if you don't have something already. Load Ubuntu from the mini.iso, no GUI, only Ubuntu Server and Samba.

      --
      "Sometimes, I think Trent just needs a cup of hot chocolate and a blankie." -Tori Amos on Nine Inch Nails
    21. Re:SMB by lysergic.acid · · Score: 1

      i'm not really a networking geek, but i've been wondering if an old 2.8 GHz P4 (with only one stick of 256MB PC133 SDRAM left working) we have sitting in the office might make a decent file server for a home office. it seems like a waste to toss the system out, but it also doesn't make sense to add more memory to the system to keep it as a workstation.

      does anyone know if such a system would perform well as a file server without a memory upgrade? or would it be possible to set it up as a wireless router? i currently have a Netgear WTG624 v3 wireless G router, but every few days i'll just start being unable to resolve new addresses or establish new connections. the problem usually goes away once the router is reset, but it's rather annoying. would an old desktop make a better wireless router?

    22. Re:SMB by JWSmythe · · Score: 2, Interesting

          I concur with this. Anything that says "GigE" only means that it's offering an interface that is compliant to the specification, not that it can pass 1000Mb/s.

          A few days ago, I went digging for some information on switches. I'm a big Cisco fan, and I have specs on everything that I use. I know which of my switches can handle more traffic than others. That's kind of important.

          Someone (to remain nameless) bought a GigE "switch". A name brand, but consumer grade switch. He wanted GigE because he had large files to transfer between several machines simultaniously.

          "switch" by their definition in the user manual simply means hub, except it can amplify the signal. No actual switching involved, other than the fact that it can "switch" between 10Mb/s, 100Mb/s and 1000Mb/s. {sigh}

          And the pps rates were pathetic. Actually, very pathetic. I broke out my spreadsheet of Cisco specs, and had to scroll down to the slowest, oldest switches that I can get my hands on. A base model Cisco Catalyst 2924 (not enterprise firmware). The 2924 handles 3 times the pps than this spiffy keen new "GigE switch". {sigh}

          I only looked into it due to other network problems. Cascaded consumer grade switches in what should be a high speed operating environment. Nothing even came close to the old Cisco 2924. While I'm not advocating running a new enterprise on old 2924's, and the fact that there are much faster ones laying around waiting for a home, wouldn't it be prudent to use something else.

          So the moral of my story.... Figure out what you're really dealing with, and don't look only at the label.

          I was having a discussion with someone who does SAN work. He was all happy about his piece of equipment. I found out the specs of the components, and then priced it out with better PC based stuff running Linux. His did run Linux, but on a custom board. It was easy to out perform anything he had with better hardware, and even better drives. If I recall correctly, he veto'd the idea of switching because he had once tried it with a Windows based SAN, and it wouldn't work. Tried once. With some 3rd party crap. It didn't work. {sigh}

          I'm slowly prepping a friends place to have a Linux machine be the SAN. Decent parts, standard protocols (SMB, NFS, and iSCSI). The only "slowly" part is that there is no rush right now, so when I see something that'll do it well, we buy the parts. Once we have all the parts, it'll be a running machine.

      --
      Serious? Seriousness is well above my pay grade.
    23. Re:SMB by Firehed · · Score: 1

      For home use? That would be more than fine. My former fileserver box had a gig only because I had it sitting around, but for only a couple of users that's more than enough. When it's accessing the same files over and over again as you may get on a busy network you'll want the extra RAM, but the raw CPU speed tends to be the limiting factor in network speed unless the system employs specialized network hardware (and the < $500 "cheap" units never do). My system was specced with... I think a 2GHz Celeron and 1GB of RAM and was pretty much limited in transfer speed by the hard drive itself - and got about the same network speeds on otherwise the same network as the drives do in their new home (a current-model Mac Pro with 10GB RAM and 8 2.8GHz Xeon cores).

      I wouldn't think that 256MB of RAM would be limiting speeds until you've got at least 25 people that are frequently hitting the box. I'm no expert on the matter, but that "feels" about right to me. Unless you decided to use Vista as the host OS (I traditionally use XP as a fileserver OS just because setting up the shares is so damn simple, but as the rest of the house/home office is OS X now, I just moved the drives).

      As for the WiFi router thing... I wouldn't bother with it. I'd replace the router, but not with a full PC. That's just overkill (at least for a home office), and is wasting power and asking for one of the many moving parts to fail. I personally use a GigE Apple Airport Extreme b/g/n and for the most part have been pretty happy with it. Never any performance issues anyways, but it (at least with the firmware I have) doesn't support uPNP, just Apple's semi-proprietary NAT-PMP which makes some apps a little unhappy. As such, I have it acting as just a Wifi repeater and gigabit switch, and let a very old D-link b/g router do the actual firewall stuff.

      --
      How are sites slashdotted when nobody reads TFAs?
    24. Re:SMB by Pikiwedia.net · · Score: 1
      I agree, SMB has good compatibility and would be a good choice.

      Consering the NAS, your requirements can be boiled down to: High performance (Stability, high throughput, large volume) Simple deployment (In terms of your own labor) SOHO pricepoint ( under $5000 )

      The problem is that you only get to choose 2.

      High performance + simple deployment = Business-grade solution installed by consultant

      High performance + SOHO pricepoint = Server with SATA Raid or ZFS

      Easy deployment + SOHO pricepoint = SOHO NAS (ReadyNAS, Synology, etc)


      Also, remember the backup. And backup is expensive, far more expensive than disk space.

    25. Re:SMB by ters+a-zA-Z0-9$_.+!* · · Score: 0

      i had an old dell running samba to my lappy on 7.10 when i did my upgrade to 8.04 samba was broken for months.I thought it was my laptop running lenny that was the problem but it was the Ubuntu samba non sense

    26. Re:SMB by dna_(c)(tm)(r) · · Score: 0

      Installing Ubuntu or whatever on it seems like more of a performance hit than a properly optimized "off the shelf" NAS box, since they most likely don't run Dbus, GNOME, Hald, bluetooth or any other desktop software atop the basic kernel and networking services.

      Just install the Server Edition instead of the Desktop Edition.

      It is already mentioned on their homepage. Server is a minimal (headless) installation...

    27. Re:SMB by MrAngryForNoReason · · Score: 1

      Be careful using SMB in an OSX environment. Unless you have policies in place to restrict the usage of special characters in file names you will find your file and folder names turn to gibberish when accessed over SMB.

    28. Re:SMB by repvik · · Score: 1

      There is no such thing as a "properly optimized" off-the-shelf NAS box. Really.
      If you want a NAS-device with a slightly optimized OS, you've gotta cook it yourself in most cases. The OS distributed with almost all NAS-devices is a huge heap of shit, thrown together as fast as possible.

      For a few devices, a distribution called "foonas" is available:
      http://foonas.org/index.php/Main_Page
      The speed difference is appreciable...

    29. Re:SMB by drsmithy · · Score: 1

      "switch" by their definition in the user manual simply means hub, except it can amplify the signal. No actual switching involved, other than the fact that it can "switch" between 10Mb/s, 100Mb/s and 1000Mb/s. {sigh}

      Wow. I didn't think it was even possible to buy a gigabit hub.

      I was having a discussion with someone who does SAN work. He was all happy about his piece of equipment. I found out the specs of the components, and then priced it out with better PC based stuff running Linux.
      [...]
      I'm slowly prepping a friends place to have a Linux machine be the SAN. Decent parts, standard protocols (SMB, NFS, and iSCSI). The only "slowly" part is that there is no rush right now, so when I see something that'll do it well, we buy the parts. Once we have all the parts, it'll be a running machine.

      Do your Linux-based "SANs" provide dual (ideally active/active) controllers with battery-backed, mirrored cache ? Because even low-end $10-$20k devices like the Dell MD3000i or EMC AX4-5 do that these days.

    30. Re:SMB by diskis · · Score: 2, Informative

      That's not Samba's fault. It's the TCP window size on XP that is the problem.
      I have at home a cheap server running Ubuntu and Samba with older drives that max out at 35-40 MB/s.
      Clients using OS X, Linux or Vista gets the full ~30 MB/s, but XP clients seem to max out at 10-15MB/s. After tweaking the TCP window size, I've gotten the speed up to 20-25MB/s.

    31. Re:SMB by yeah-whatever · · Score: 1

      11MB/sec sounds like a saturated 100Mbit network connection. The fact the the 2 PCs are able to achieve more suggests that the network connection between them is 1000Mbit all the way. I would suggest you have a close look at all the network interfaces between you and the SAMBA server. It's likely one of them is 100Mbit

    32. Re:SMB by Anonymous Coward · · Score: 0

      maybe it's the slow virus-scanner on the windows
      box? just for testing: disable any anti-virus on XP
      for the speed test ...

    33. Re:SMB by LarsG · · Score: 1

      Wow. I didn't think it was even possible to buy a gigabit hub.

      Half Duplex is part of the GigE spec as far as I know, but I've never seen a GigE hub. WJSmythe, could you share the manufacturer and model number on this oddity?

      --
      If J.K.R wrote Windows: Puteulanus fenestra mortalis!
    34. Re:SMB by LarsG · · Score: 1

      I cant get it going faster than 10-11MB/sec when copying to/from Windows XP.

      It could be something as simple as the network card in the server autoconfiguring to 100Mbps, 11MBps sounds like a saturated 100Mbps link. Check with ethtool on the server and in the management interface on the switch. Bad performance like this can also be caused by mismatched duplex on the server network card and the switch it is connected to.

      --
      If J.K.R wrote Windows: Puteulanus fenestra mortalis!
    35. Re:SMB by Anonymous Coward · · Score: 0

      He did mention that he was using a 100Mb connection.
      The max theoretical throughput would be 12.5MB/Sec.

      45-60MB/Sec is pretty slow for a 1000Mb connection though, you should be able to push around 100MB/Sec without any optimisation as the max theoretical throughput on a GigE line is around 125MB/Sec.

      (Obviously you'll never actually be able to hit those max speeds for any real life situation as there will be overhead involved with any protocol you use to transmit data over the line.)

    36. Re:SMB by 1310nm · · Score: 1

      Is it bad that I automatically assume I'm being pranked now when someone posts a tinyurl?

    37. Re:SMB by LWATCDR · · Score: 1

      Well if you install Ubuntu server or CentOS then you have the option to do just a server install. No X, Gnome, Bluetooth, DBus or other fluff. FreeNAS and OpenFiler are NAS specific distos so they are already stripped down and pretty dang light. So that combined with a custom box is pretty much a NAS unit.
      The key thing here is that he really cares about performance. You must pay for performance. The thing is that there are two ways to pay. Money or time. Putting together a NAS box will take a little time but depending on how much free time he has at work it will be cheaper than buying a high end NAS box.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    38. Re:SMB by LWATCDR · · Score: 1

      Would work great with FreeNAS. As to using it as a Wireless router? I am not sure about that.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    39. Re:SMB by ckaminski · · Score: 1

      All of which can be turned off, leaving more CPU and memory for shifting packets, and doing crazy stuff like iSCSI.

      I've looked at all the NAS devices on the market, and for the quantities of data our 1.5TB drives offer, they're all horrid. Better off spending a few more dollars on a general purpose PC and a small case, and put a version of FreeNAS or Openfiler on it.

      A) better software support
      B) open technologies
      C) faster, better, not necessarily cheaper
      D) as many drives as you can fit in whatever case you buy.

    40. Re:SMB by Anonymous Coward · · Score: 0

      "Wow. I didn't think it was even possible to buy a gigabit hub."

      It is not. For various technical reasons, gigabit Ethernet *must* be switched. If the manual says otherwise it's probably a bad Engrish translation.

      To the grandparent poster - Did the user of this noname "switch" get increased performance with a different switch, or was it "bad performance, blame the switch because of the manual"? Did it even perform badly at all or are you just saying "the specs don't match a Cisci therefore it'll have to perform awfully!"

      99% of the time, performance problems are due to the endpoints and not the switch itself. The exceptions are extreme corner cases such as lots of tiny packets (but this almost always bogs down the endpoints first) and "smart" switches that are doing some sort of per-packet processing and QoS that have somehow been misconfigured such that hardware is no longer accelerating functionality. BTW the latter performance-killer can even happen with Cisco hardware.

      In my experience, except with extremely high end endpoint hardware or a large network (not the GP's friends' small home network), you will never hit a pps limit on the switch before you hit a pps limit on the endpoint.

    41. Re:SMB by Anonymous Coward · · Score: 0

      We recently retired an old Netware box, serving files to 25 users. It had a PII 350 & 256mb RAM, which was just fine. It was running Seagate Cheetah drives; for a simple fileserver I/O is more important than anything else.

    42. Re:SMB by fubar1971 · · Score: 1

      I currently run a small online backup business. I put an external storage device in and connected it to a server running FreeNAS. The data through put to the storage is extremely fast. My main bottle neck is my customers slow SMB or residential internet connection, so my switches are not the fastest and don't really need to be. Where I spent my time and money was the storage. Of course if I need to move the data internally then the switches would need to be looked at. For my situation this works well.

      From my limited experiences and what the parent post is looking at, IMHO, to implement a respectable but cost effective SANs, spend the money on enterprise level switches and fast drives, and then save a couple of bucks by implementing FreeNAS connected to an externally attached storage device that contains your fast drives.

    43. Re:SMB by fubar1971 · · Score: 1

      If you really want to go cheap, but have more than enough CPU power, buy a used poweredge with dual PIII 1 GHZ processors and install FreeNas. I did this when I was still experimenting when setting up my online backup business. Worked great! I have since replaced the FreeNAS hardware with a newer server (for reliability, I don't know how many ours were on the used one), but still keep the used one around as a spare. The actual data throughput between a new low end poweredge and the used one with PIII 1GHz processors was non existent.

    44. Re:SMB by fubar1971 · · Score: 1

      "nstalling Ubuntu or whatever on it seems like more of a performance hit than a properly optimized "off the shelf" NAS box..."

      This is true. Take a look at FreeNas. FreeNas is based on FreeBSD and is optimized to run as a NAS server. That is it's only purpose. Works great.

    45. Re:SMB by Big+Boss · · Score: 1

      I have OSX using NFS shares right now. They work far better than SMB or AFP did with the automounter that comes with OSX. It required a little work on the OSX box to get the UID/GID situation set up properly so security works, but it wasn't really that hard to do.

      OSX doesn't deal with sleep and such very well with the automounter and SMB/AFP. NFS seems to work just fine though. Interesting, I would have thought AFP would be better. The Linux fileserver is running all three protocols and they all work well. I find I use NFS the most as I have it serve movies and such to the Myth systems.

    46. Re:SMB by koafc2 · · Score: 0

      Tomshardware reviewed SS4200-E and it looks significantly better than you describe. http://www.tomshardware.com/reviews/ss4200-e-nas-raid,2076-5.html

    47. Re:SMB by JWSmythe · · Score: 1

          The user believed he had increased performance, because his switch said "GigE" on it. He probably did see some performance increase, because he was using a white box 10/100 hub before that. The users are complaining a lot about performance though. Inside the office, we have a lot of machines. He moved a server that I'm responsible for over to one of the cheap "switches". I was getting 0.28ms. I was also having the network hang mysteriously pretty frequently when people are working on other machines attached to those "switches". I moved the connection over to our core switch, a Cisco Catalyst 5500. The latency went down to something like 0.14ms, and the mysterious hanging went away. I don't do large transfers or anything particularly intensive from my desktop to it. I simply ssh to it. It's not so much fun typing out an entire line, and waiting for it to show. :)

          I do agree with the fact that most home users don't need anything better than the regular consumer grade stuff. Really, at home between my own computers, I use whatever crap is cheapest. :) Knowing that I will never exceed by upstream bandwidth, and all my computers only talk out over the internet and not very much between each other, I personally will rarely see a use for good hardware at home. I do have Cisco gear at home, but it's to use as spares for real production environments, or to sell on eBay. :) Heck, I have a Cisco 5000 switch sitting in the back of my car right now. That's way overkill for the 20Mb/s upstream bandwidth that I have.

          I know Cisco can get bogged down with tiny packets. That's why you have to keep your network up with the proper PPS rates for your network.

          I do know someone who could really use good Cisco hardware. He runs virtualized environments. That is, he's made a business out of putting people up on VMWare servers. He serves all of their stored data over a SAN, and that makes a lot of load on his switch all the time. He's using iSCSI, not fiber. He wants to upgrade from his HP switch to a Cisco, but doesn't have the budget for a good Cisco with GigE on all ports. I've seen his bandwidth graphs. They'll pretty regularly go over 200Mb/s, so he actually needs the bandwidth.

          For the other users I mentioned, they'll only very very rarely get up to 80Mb/s, because of limitations in what they're doing.

      --
      Serious? Seriousness is well above my pay grade.
    48. Re:SMB by greg1104 · · Score: 1

      Particularly with cache, storage vendors across the board are offensively stingy (16 gigs of high-quality ECC RAM costs maybe $800, but you'll be lucky if your $100k SAN comes with half that amount).

      Sure, but a good SAN has to worry about not losing that cache in case of a power failure, which adds a series of costs and design trade-offs that drive the cost up beyond just the RAM. If you've got an application that wrote and then fsync'd the write, the SAN will say it's committed to the OS once it reaches the write cache. That type of write caching is essential for accelerating database commits for example. But once you've done that, you cannot lose the results, as some applications will end up with inconsistency that ends up corrupting the file if that the writes it did before the fsync are partially committed. So a SAN write cache needs to worry about powering the RAM from a battery if there's a failure, and the good ones will set aside a chunk of a disk just for the purpose of dumping this data there in a hurry for an emergency shutdown. (This can be much, much faster than writing it to the actual disks because the cache can be filled with random writes while the dedicated save area is a contiguous block).

      That's one reason why it's not really fair to compare a chunk of RAM to an enterprise grade write caching solution. Sort off topic for the purpose of a home server, unless you want to run a database on one of its mount points that is.

    49. Re:SMB by LoRdTAW · · Score: 1

      I know this article is old but I want to thank everyone for their help. The few who suggested I run ethtool were right on the money. I ran it and lo and behold it was running at only 100mbps! I traced the problem to a bad ethernet cable. I am red-green color blind and the link lights on the switch change color (red or green) according to the link speed. And wouldn't you know it I was seeing the wrong color! My brother said "dude that light is as red as blood". I thought it was green, DOH!

      Thank you all again! I now get over 40 megabytes a second to windows XP and Vista!

    50. Re:SMB by greg1104 · · Score: 1

      That will work fine for a server, as long as you have reasonable expectations for the result. I used to have a server with extremely similar specs here running Linux as a file server, and it's only recently I retired it. I was hard pressed to get much over 200Mbps out of it over the PCI gigabit card I put in there. That's at the low side of real gigabit results, but completely acceptable for most purposes.

      I've had nothing but trouble using a regular server OS as a wireless router though. Wasted lots of time trying to get stuff like VPNs and torrents working and similarly tweaking the firewall rules. Nowadays I just buy cheap routers that run DD-WRT instead. For example, the same Linksys router I have that used to have weird problems all the time like you describe (needed regular resets just to work normally) is rock solid running the DD-WRT firmware instead--uptime in months instead of days.

    51. Re:SMB by speculatrix · · Score: 1

      most of the procurve medium end stuff is wire-speed gigabit; we just dropped GB£5300 (about US$8000) on a 5412zl - a bundle deal of chassis + dual PSUs plus 96 ports giga switching with PoE option. You can't get modular/chassis cisco stuff of that capability for anything like that price, and the cisco stuff is very power hungry too.

      that said, I prefer cisco configuration by far to procurve; we have some very nice 3650E's which provide very impressive performance with the 10G uplinks!

    52. Re:SMB by joib · · Score: 2, Informative

      Unfortunately Using Samba is almost 10 years old by now, and some of the tuning advice might not be applicable any more. In particular, newer versions of the linux kernel (2.6.17+) have full tcp autotuning. But explicitly specifying buffer sizes (socket options SO_RCVBUF and SO_SNDBUF) will disable this autotuning. So using some value that was good 10 years ago (8192) might be pretty far from optimal these days.

    53. Re:SMB by unitron · · Score: 4, Funny

      The user believed he had increased performance, because his switch said "GigE" on it

      Does his Cat 6 say "Monster Cable"?

      --

      I see even classic Slashdot is now pretty much unusable on dial up anymore.

    54. Re:SMB by JWSmythe · · Score: 1

          hehe. That was so mean. :)

      --
      Serious? Seriousness is well above my pay grade.
    55. Re:SMB by drsmithy · · Score: 1

      Sure, but a good SAN has to worry about not losing that cache in case of a power failure, which adds a series of costs and design trade-offs that drive the cost up beyond just the RAM.

      Battery-backed cache and cache destaging are hardly a groundbreaking ideas, nor is scaling the battery or "UPS" capacity to allow for the cache size.

      That's one reason why it's not really fair to compare a chunk of RAM to an enterprise grade write caching solution.

      It most certainly *is* fair. Larger amounts of cache available on higher-end systems (that are otherwise using the same theory - eg: EMC AX150 vs CX3-80) demonstrate the technical aspect(s) of the problem is/are already solved. At which point the issue *is* just the marginal cost of more RAM (and possibly more battery capacity).

    56. Re:SMB by Somegeek · · Score: 1

      FreeNas is not supposed to be a production ready product. Note this statement on their site:

      http://www.freenas.org/index.php?option=com_openwiki&Itemid=30&id=faqs:en

      "FreeNAS is Alpha or Beta, its not a production release and it will have bugs in it. It is your risk if you load valuable data onto FreeNAS."

      Why you would risk your customer's data, or recommend that others do the same, is beyond me.

      --
      And as you tread the halls of sanity, You feel so glad to be, Unable to go beyond. I have a message, From another time..
    57. Re:SMB by fubar1971 · · Score: 1

      This is true, but if you test it and run it through its paces first, it is a risk that is minimal. I tested it for almost a year before I put it into production. Also if you plan properly, you reduce the risk even further. If the FreeNAS server does fail or experiences a bug, I can always put in place a different server to access the storage device since it is external and can be accessed using open standards. Just because it is beta doesn't mean it will not work. Not to start a flame war, but Look at WinBlows. M$ has been pumping out beta and alpha quality products and charging people for it.

  2. You could roll your own. by LWATCDR · · Score: 4, Informative

    FreeNAS or OpenFiler on a PC with a raid controller and GigE should work. It might even be cheaper than a NAS box.
    As to OS/X support. I thought OS/X supported Windows networks out of the box. Odds are very good that if it supports Windows OS/X will work.

    --
    See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    1. Re:You could roll your own. by sizzlinkitty · · Score: 1

      I agree, out of the box solutions never work great and are not really dependable. My vote goes to FreeNAS OR a normal Debian installation. If network throughput is a issue, look into bonding the adapters going back to the switch.

    2. Re:You could roll your own. by nhtshot · · Score: 5, Informative

      My situation is similar to yours. I bought and tested several off the shelf solutions and was continuously disappointed.

      My solution was an off the shelf AMD PC filled with HDD's and linux software raid.

      It's MUCH Faster (90MB/Sec) then any of the NAS solutions I tested.

      With Christmas specials abounding right now, HDD's are cheap. Use independent controllers for each port and a reasonable CPU. Also make sure that the GIGe interface is PCI-E.

    3. Re:You could roll your own. by Anonymous Coward · · Score: 0, Flamebait

      Having to terminal in to it in a time of need and reconfigure it, or solve package dependencies on it is probably not what the poster had in mind. Speaking of which, since this is not a story why isn't this ask slashdot?

    4. Re:You could roll your own. by nine-times · · Score: 1

      I thought OS/X supported Windows networks out of the box. Odds are very good that if it supports Windows OS/X will work.

      Yes, OSX supports SMB via Samba, which means it has solid support for Windows file sharing. You can run AFP on Linux or Windows, but frankly it's not really worth it. I'd be interested to know if anyone wants to make a case that AFP is necessary, but my personal opinion is that it's only worth using if you're running an OSX server.

    5. Re:You could roll your own. by AngelofDeath-02 · · Score: 2, Informative

      Sadly, my off the shelf pc is woefully insufficient ... I get 24MB/s max from a raid over gigabit ...

      The pc was originally an AMD 1800+ with SDRAM.

      there are 8 drives total, one boot (80 gig IDE)

      7 250 gig Seagates, all IDE - Originally they were all on a separate controller, and I used a raid controller to do it (acting as ide, no raid, in this case) the 7 250 gigs are setup in a software raid5 configuration in linux. Individually hdparm rates them as 60MB/s, and the whole raid as 70MB/s but for whatever reason file transfers from the raid to the boot drive topped out at 20-30MB/s. The gigabit card was also on the same PCI bus. However, copying from the boot drive over the network went at 50MB/s.

      Thinking it might be a PCI bus limitation somehow, I moved the raid into a newer motherboard, sporting a 2200+ and ddr memory. Now being limited in ide controllers, all 7 drives are plugged into the raid controller in a master/slave setup. I get similar performance (average is now 23MB/s vs 20) and the gigabit ethernet controller is onboard.

      I can't figure it out -_- I also don't have the money for a second raid controller (to put them all on their own channel) or to rebuild the pc with a PCI-E bus and sata.... so for now, that 20MB/s will have to be sufficient.

      --
      No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
    6. Re:You could roll your own. by haemish · · Score: 1

      I've had phenomenal success with OpenSolaris. ZFS is the coolest way to run a whitebox JBOD. The box I built can easily drive a GigE to saturation.

    7. Re:You could roll your own. by syntax · · Score: 1

      I haven't directly diagnosed this issue since 10.3, but it still might be an issue:

      OSX does support SMB pretty well (they actually use the samba suite under the hood for client and server). There's a catch though. In MacOS (classic and X), there are two parts to the file: the "data fork" (what you would normally think of as the file), and the "resource fork" (contains meta data, and executable code for "classic" programs). Over SMB, the resource forks are stored as a separate file; example.txt's resource fork would be stored as ._example.txt.

      This by itself is not a problem. The problem is that many OSX programs will lock the resource fork, but never unlock it. AFP never has an issue with this (I assume its built into the protocol, as it is into HFS). The net result is that if you're running multiple users trying to successively modify a file, they're going to get locking errors when they go to save.

      I ran into this issue trying to maintain an all Linux development environment for a team of web developers. In the end I had to just get a dedicated OSX box so all the mac clients could work happily. Also, netatalk was an option, but at the time it was too immature, and did things like enforce a small (15?) character filename limit.

    8. Re:You could roll your own. by bu1137 · · Score: 1

      10.4 at least also has a few problems with smb. Locking for example, and a bug with aliases, as far as i know. 10.5 seems to have fixed those two problems i've experienced.

    9. Re:You could roll your own. by jbwiv · · Score: 3, Informative

      The problem with that is power consumption. Build your own, and you'll be burning a lot of power unnecessarily because it's overkill. Contrast that with the ReadyNas Duo, which I own, that pulls on average around 30-40W. Much better and green.

    10. Re:You could roll your own. by Luthair · · Score: 1

      One consideration with a PC is that they will likely require significantly more power than a NAS box, even if you buy a low power system.

    11. Re:You could roll your own. by bu1137 · · Score: 2, Informative

      Get yourself an AMD 64 X2 4850BE (2.5 Ghz, 45W), a mainboard with AMD 780G chipset, an efficient power supply and two western digital green power drives. That'll eat about 40 watts on idle.

    12. Re:You could roll your own. by Anonymous Coward · · Score: 0

      i use an RMD2150 for $500 it saturates our network link.
      runs linux to boot.
      opensolaris is overkill. smallnetbuilder regularly tests out NAT boxes...he should look at the smallnetbuilder website.

    13. Re:You could roll your own. by ceoyoyo · · Score: 3, Informative

      Most NAS devices, particularly the consumer ones, cheap out on the processor. You might have great hard drive throughput, maybe even a nice fast network interface, but the poor little processor just can't keep up.

      If you want speed, definitely throw a PC at it.

    14. Re:You could roll your own. by Score+Whore · · Score: 3, Informative

      You have seven drives in a software raid5. Anytime you do a write, the entire stripe has to be available to recompute parity. If you aren't doing full stripe writes, that will often mean having to read data in from a portion of the drives. A normal PCI slot will give you 132 MB/s max. Possibly that is a limitation, but it's higher than gigabit speeds so you may not care that much. Also your raid controller may not exactly be lightning. But I'd personally suspect the number of columns in your RAID5.

      Also, as a little learning experiment, take a drive, make two partitions of a few gig each. Put one of them at the beginning of the drive and put the other at then end of the drive. Benchmark the speed of those two partitions. In case you're not really that interested, the laws of physical make the bits at the outer edge of the platter go by about twice as quickly as the inner edge. So if you are doing a sequential benchmark you'll find that a disk that rates 60MB/s on the outer edge will drop to 35MB/s on the inner edge. So on average, you'll find that the majority of your disk isn't as fast as simple sequential tests suggest.

    15. Re:You could roll your own. by jvin248 · · Score: 1

      FreeNAS would be a good choice - easy to install and run. You could build a Debian server that might be faster but wouldn't have the easy interface that FreeNAS has (I use debian based distros so recovery is easier than pulling files out of a BSD based FreeNAS).

      Ignore the issues about energy use with a pc. You can probably choose an older PC, set the energy management features, and be close. I've had towers that run just 30w actual (using a meter) where most people read the power supply label that says "300w" and freak out that their build uses huge amounts of energy all the time.

      Read the FreeNAS forums (or post your question) as some have done extensive and impressive speed runs with their builds.

    16. Re:You could roll your own. by gad_zuki! · · Score: 1

      Really? My ReadyNas 4TB was pulling 60-80W on idle.

      FWIW, only could get 25/MBs with SMB/CIFS on a gigabit network.

    17. Re:You could roll your own. by Anonymous Coward · · Score: 0

      At my office, we build a solution like that for customers wanting a low end solution for HA/DR storage for VMware ESX. Toss together any modern dual/quadcore proc, shove an Areca (we've had issues with 3Ware, so I steer clear of them) RAID controller in there, and start stacking the SATA hard drives in.

      If you're really brave, skip something like OpenFiler and set up a vanilla server distro with LVM, then expose the volumes through AOE on a separate SAN to get some real speed going. We've run 5 virtual servers through iSCSI on a RAID 6 of 8 drives and had no problems.

    18. Re:You could roll your own. by UU7 · · Score: 1

      OpenSolaris 2008.11 runs great, the timeslider is a very useful feature as well.
      zfs is quite easy and more reliable than a raid 5 setup.

      http://breden.org.uk/2008/03/02/a-home-fileserver-using-zfs/ is a great guide to help you out. I get 45MB/s read and write from a dual parity array with 4 640gb drives. The cpu isn't the most powerful but it gets the job done.

    19. Re:You could roll your own. by LWATCDR · · Score: 1

      Maybe but AMDs 780G chipset combined with one of the BE cpus will give have a pretty low power draw. Setup the power management and it will not be much more than 40 watts ideal. The key thing here is most NAS boxes will come no where near saturating a GigE connection. If you need the power you need the power. If not then yes a NAS box might be a better choice.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    20. Re:You could roll your own. by LWATCDR · · Score: 1

      Ummm... You haven't see FreeNas or OpenFiler have you.
      They are actually very robust and complete. Plus for a NAS once you set it up you not going to do a lot of reconfiguring or adding packages. With OpenFiler you may want to run YUM every now and again to get the latest updates but even that is optional.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    21. Re:You could roll your own. by PiSkyHi · · Score: 1

      With 4 drives on Pentium 4 3 Ghz, RAID 5 takes around 5% to 10% of the CPU when I saturate the Gbit ethernet.

      Independent controllers is a good choice - I chose a basic PCI-E 2 port SATA card with a SiI 3132 - 2 of them. Under Linux, drives are hot pluggable and software RAID suddenly becomes superior to hardware RAID.

      Also, I would say use at least 4 drives if you want to use the server for anything more like a small webserver or something - saturating the drives instead of the gigabit will cause multiple access to really slow down. Having 4 drives ensures local array read performance - especially latency, is still excellent.

    22. Re:You could roll your own. by rthille · · Score: 1

      What issues were you seeing with the 3Ware cards? I've got a friend with a 16-port he loves...that server replaced an older one with an 8-port (IIRC). He did see some performance issues with his setup, and 3Ware worked with him and they figured out that it had to do with queue depths for the I/O's and the fix was a simple change to his config.

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    23. Re:You could roll your own. by LWATCDR · · Score: 1

      I have to admit that I have thought that a Solaris based FreeNAS or OpenFiler could be very interesting.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    24. Re:You could roll your own. by blincoln · · Score: 3, Informative

      I'd be interested to know if anyone wants to make a case that AFP is necessary, but my personal opinion is that it's only worth using if you're running an OSX server.

      Our Mac people at work claim that the only way for the OS X file search utility to work correctly is via AFP. The third-party server software they use as an AFP server on Windows maintains a server-side index, which I imagine is why, although I don't know how much of that is a requirement with OS X as opposed to their specific configuration.

      --
      "...always new atoms but always doing the same dance, remembering what the dance was yesterday." -Richard Feynman
    25. Re:You could roll your own. by curmudgeous · · Score: 1

      I strongly suspect it's IDE that is killing you as Master/Slave drives on the same cable can't both operate at the same time (they take turns). Dump the IDE and replace the drives with SATA. You can pick up 4-port Promise SATA controllers for cheap (around $50/ea) and just do software RAID 5.

      My home file server is made mostly from scraps: 5 x 250 GB SATA II drives (software RAID 5); two 4-port Promise non-RAID controllers; an Athlon XP 3000+ CPU; 1 GB RAM. I use an old IDE drive loaded with OpenSUSE 11 for the boot volume. The data volume is exported as an SMB share on a gigabit network. I get on average 1.5 GB/min transfer rates across the network.

    26. Re:You could roll your own. by styrotech · · Score: 1

      The difference in power makes sense. The disks are what use the majority of the power in these units. You had a 4 disk unit, they had a 2 disk unit so yours used more power.

      As for idle power on the ReadyNAS - some of the desirable disk access options for either performance or reliability (I forget which) are incompatible with spinning down the disks, so the disks can still eat a fair chunk of power even when the unit is idle.

    27. Re:You could roll your own. by Anonymous Coward · · Score: 0

      I second that, FreeNAS rocks. Has anyone played with rsync on FreeNAS, im looking to setup a NAS and have a simultaneous mirror at another location.

    28. Re:You could roll your own. by htnprm · · Score: 1

      Uh. Let me report my experiences with my Netgear ReadyNAS NV+. Pull it out of the box. Turn it on. Some basic config to get it talking to Active Directory. Copy data. Set up some shares. Check ACLs. Rock and roll. Had a single drive failure at one stage. Simply replaced the drive.

      Prior to this I was hosting the files on a Windows Server 2003 SBS box. Oh look. A new patch for Notepad *reboot*. And don't try to tell me the problem was I was using Windows. I restart my Ubuntu workstation fairly frequently as well. The NAS has been restarted twice in eight months (Both times because it was being moved).

    29. Re:You could roll your own. by edmudama · · Score: 2, Interesting

      The OP stated they have a business need for moving gigabytes of data quickly around the office. Spending the extra money on a real server's power consumption would save them thousands of dollars a day worth of their time.

      Even the cost of the power is pretty minimal for this... Figure 500 watts for 24 hours is 12 kWh. At worst you pay $0.20/kWh, which is a hair over $2/day, assuming 24 hours/day usage. My linux PC NAS in the basement saturates gigE and is under 100 watts active power consumption, or about $0.40/day by california pricing.

      I think those little mini NAS boxes suck for all but the simplest home applications, and they're penny-wise/pound-foolish for data-intensive business applications.

      --
      More data, damnit!
    30. Re:You could roll your own. by Psiven · · Score: 1

      This has been fixed since Leopard. It was a major pain in the ass though.

    31. Re:You could roll your own. by Anonymous Coward · · Score: 0

      And don't try to tell me the problem was I was using Windows. I restart my Ubuntu workstation fairly frequently as well

      Well if it isn't Windows, or Linux, it must be... you? I've used debian on various cheap second hand hardware (i486, dec alpha, pentiums, athlons, celerons, via C3) as pos home servers like forever and hardly ever had to reboot them.

      Btw it looks like you forgot "mount it in a safe place". Safe being: the nas box can't drop, you can't trip over cables, can't easily spill liquid on it. If you're living below sea level like me and moving your NAS up a floor would get it above sea level... well it'd be nice to at least have your backup data if there's ever going to be a flood.

    32. Re:You could roll your own. by dimension6 · · Score: 1

      I also went that route for two reasons. A dedicated NAS box such as the ReadyNas Duo not only uses less power than an old PC, but also takes up much less space. This is an important point for me, as I live in central Tokyo (read: tiny apartment).

    33. Re:You could roll your own. by Sentry21 · · Score: 1

      Even better, OS X supports NFS, which is much faster in every test I've tried (from random access times to raw throughput), likely due to significantly lower overhead due to a simpler protocol.

      Export the same shares out using NFS and SMB and you're golden.

    34. Re:You could roll your own. by ischorr · · Score: 2, Informative

      OS X doesn't support the ability to CHANGE CIFS (SMB) permissions, so that's a concern. It can at least change NFS permissions, if only from the CLI or other Unix permissions-aware apps.

    35. Re:You could roll your own. by Anonymous Coward · · Score: 0

      Over cifs it has to scan each directory across the network, AFP lets them use the spotlight indexes. However I'd imagine you'd need an OS X machine to build said indexes.

    36. Re:You could roll your own. by Anonymous Coward · · Score: 0

      The issue is that SMB on OSX only supports the smaller FAT numbers so your maximum file and volume sizes are limited.
      However, if it supports AFP there's no problem.

    37. Re:You could roll your own. by Anonymous Coward · · Score: 0

      Don't bother with gigabit with freenas, it can barely keep up with 10Base-T, let alone 100Base-T. Openfiler costs money if you want it to actually work.

    38. Re:You could roll your own. by bcmm · · Score: 1

      If you roll your own NAS, just use SSH and rsync. Then it will support everything, and, in my experience, have more reliable transfers.

      --
      # cat /dev/mem | strings | grep -i llama
      Damn, my RAM is full of llamas.
    39. Re:You could roll your own. by Anonymous Coward · · Score: 0

      You can run AFP on Linux or Windows, but frankly it's not really worth it. I'd be interested to know if anyone wants to make a case that AFP is necessary, but my personal opinion is that it's only worth using if you're running an OSX server.

      Around here, firewall rules block SMB between subnets to minimize the various windows worms. AFP lets us share files without creating a bunch of host-specific firewall rules.

    40. Re:You could roll your own. by 1310nm · · Score: 1

      Remember, though, that FreeNAS is still not stable software. I had trouble with it not remembering settings and a few other things. Great concept, but I can't wait for them to iron out a few wrinkles.

      After reading horrible reviews about set-top NAS boxes, I built the client a PC and threw Debian and samba on it. I use Acronis True Image Home on their PCs, and have it set to use no more than 100GB of share storage for backups. It worked great for a while, but then there was a power issue that zapped the power supply and got most of the components in the machine, and now I'm trying to get it all working right again.

      In retrospect, if I had it to do all over again, I would have gone with a cheap Dell server with a warranty or something rather than building the machine myself. I believe the extremely cheap components I used were a contributing factor in the failure. Which leads me to another lesson learned: always talk to the customer about their price expectations! Don't always assume they want it as cheap as possible! In my case, I talked the customer out of manually backing up to a USB drive with a few computers into instead using imaging software to make backups to this NAS. Therefore, I assumed they would be shocked at a high sticker for the PC used as a NAS, when comparing it to the cost of a USB drive. Later, they made a comment about cost along the lines of, "Don't worry about how much it's going to cost, let's just get it working." That kind of threw me for a loop.

      On the plus side, such a configuration can be used to add value later, and the customer has asked for a light website that would provide their customers with the ability to upload files that are too big for email. The server should be able to handle that small increase in activity just fine.

      Keep in mind that most of the small business/home office sites don't have incredibly stable power infrastructure. Make sure that the NAS has a UPS, and a daemon running to monitor it and shut down if necessary. In my case, it's an APC 800VA UPS and the NAS server runs apcupsd.

    41. Re:You could roll your own. by frinkster · · Score: 1

      Yes, OSX supports SMB via Samba, which means it has solid support for Windows file sharing. You can run AFP on Linux or Windows, but frankly it's not really worth it. I'd be interested to know if anyone wants to make a case that AFP is necessary, but my personal opinion is that it's only worth using if you're running an OSX server.

      I have an old Ubuntu box at home acting as a file server on my network. I have a Macbook and my wife has a Windows XP laptop. I set it up so I connect to the server via AFP and she connects via SMB. File transfers are definitely faster over AFP, with the added bonus of the shared device automatically showing up in the finder.

      You can share the file system via SMB and AFP at the same time and both are very easy to set up on Linux (I used simple tutorials I found online).

      If you've got Macs on your network it seems like a no-brainer to enable AFP.

    42. Re:You could roll your own. by nine-times · · Score: 1

      Oh, I didn't mean to say it was hard to do, it just seems like it's not necessarily worth doing. When I've done it in the past, there were enough problems to turn me off of it. I ended up seeing files through SMB that were there to hold the resource forks in SMB, and sometimes it seemed to allow for different file naming conventions, which caused me some problems.

      Now maybe I was just doing something wrong, but I've never had great luck with netatalk or with Linux's HFS support.

      I forget the details, but SMB works well enough, and then at least OSX knows it's dealing with a non-Apple server. Except for creating some hidden files, it generally plays nice with others. If there's a big performance improvement in AFP, I never saw it in my limited testing.

    43. Re:You could roll your own. by TheRaven64 · · Score: 1

      NexentaStor is an OpenSolaris-based NAS distribution, but it's non-Free (and non-free). Fortunately, ZFS in FreeBSD is rapidly progressing towards stable, so FreeNAS should get support for it sometime over the next year. Until then, there's RAID-3 in GEOM.

      --
      I am TheRaven on Soylent News
    44. Re:You could roll your own. by TheRaven64 · · Score: 1

      $0.40/day is $146/year, which is a fair amount. Personally, I'd like a NAS that consumed something closer to 1W when idle. It's a shame there aren't any ARM-based boards available for this kind of thing. The BeagleBoard draws 1.8W for the entire board, and the DSP should be fast enough for offloading ZFS checksum calculations. Unfortunately, it doesn't have any way of plugging in hard disks. If it did, and could power them down when not in use, then it would be ideal.

      --
      I am TheRaven on Soylent News
    45. Re:You could roll your own. by jp10558 · · Score: 1

      But if they're moving what, 100s of GB or more a day, I'm thinking they might want more than 1.5TB of storage, which means more disks. Disks take up space AND power. I can't imagine getting much smaller than my mini-tower box with 9 disks in a RaidZ2 array + CD Rom for booting/installing/patching and one OS drive. 10 3.5 inch disks take up room...

      Granted I built this with 750GB disks so only get about 4.5TB of available space according to Windows via CIFS fileshare, but hey.

      At home, problems are noise and heat. At work? A server room would take care of this, but they might not have one, in which case, put it away from people's desks (it's not THAT loud, but noticable if right next to it).

      --
      Opera, Proxomitron-Grypen,GPG 0x0A1C6EE3
    46. Re:You could roll your own. by LWATCDR · · Score: 1

      I just built my wife an AMD 780G system. It is a really nice chip set. The graphics are actually pretty good and it had six SATA ports.

      I wounder if you could use the GPU on that board to off load checksums and maybe encryption? What a nice NAS system that would make.

      --
      See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
    47. Re:You could roll your own. by WuphonsReach · · Score: 1

      7 250 gig Seagates, all IDE

      RAID-10 would have been a better choice for performance. Six drives for the array, with one hot-spare on standby. Net space would only be around 700GB, but you'd get pretty good performance.

      --
      Wolde you bothe eate your cake, and have your cake?
    48. Re:You could roll your own. by WuphonsReach · · Score: 1

      The problem with that is power consumption. Build your own, and you'll be burning a lot of power unnecessarily because it's overkill. Contrast that with the ReadyNas Duo, which I own, that pulls on average around 30-40W. Much better and green.

      The modern AMD 45nm energy-efficient parts have a max TDP of 45W. Which means that they're going to run at a lot less then that when not under load. Couple that with the new 500GB SATA 2.5" laptop drives and you're looking at a regular desktop tower with similar power requirements as the ReadyNAS. Or at least something in the ballpark of 30-60W.

      As much as I like the ReadyNAS units (which are Linux-based and in an pinch data can be recovered using a Linux machine), I don't care much for the proprietary power-supply or other parts that I can't pickup at the neighborhood computer parts store.

      (I've got a tits-up ReadyNAS NV sitting here beside me. Motherboard is completely fried, probably due to a defective power-supply. The unit is out of warranty, since it was bought before Infrant got bought up by NETGEAR. Waiting on the client to decide whether to spend $950 for a new unit, or whether they want us to simply build a whitebox PC using one of the NAS-like distros mentioned here.)

      --
      Wolde you bothe eate your cake, and have your cake?
    49. Re:You could roll your own. by AngelofDeath-02 · · Score: 1

      Storage space was my primary concern - I just wanted enough redundancy to survive a drive failure. I have 1.3tb usable space and it's almost full, so raid 10 wouldn't cut it =/

      In terms of priority, Space>redundancy>speed. I'm just dissapointed that it's actually slower than a single disk - by a lot.

      --
      No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
    50. Re:You could roll your own. by SETIGuy · · Score: 1

      I'd be interested to know if anyone wants to make a case that AFP is necessary, but my personal opinion is that it's only worth using if you're running an OSX server.

      I can personally make a case for it. My home network contains no Macs new enough to run OSX and includes a Mac SE and an Apple IIgs that talk through a localtalkethertalk gateway.

      Now you could claim that I'm an idiot or a masochist for wanting my IIgs to talk to my NAS box. But if I shut down the SE and the IIgs, my IBM XT would be the only machine on the localtalk network.... And that just wouldn't be right.

    51. Re:You could roll your own. by nine-times · · Score: 1

      You got me. That's a good reason to use AFP.

    52. Re:You could roll your own. by bill_mcgonigle · · Score: 1

      I'd be interested to know if anyone wants to make a case that AFP is necessary, but my personal opinion is that it's only worth using if you're running an OSX server.

      You'll eventually run into file naming issues with SMB. AFP deals better. netatalk's AFP/TCP support is plenty fine, though you'll need to db_recover its berkeley files when you do OS upgrades.

      Also some apps don't deal with the foreign file system support for some reason. It might be better on Leopard, its copyfile() is better. Leopard also has better NFS support, I've been meaning to try that. Maybe 10.5.6 is stable, 10.5.x :: x =5 have all been majorly buggy for me so I've only read about it, but supposedly the automounter at least is derived directly from OpenSolaris.

      --
      My God, it's Full of Source!
      OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
    53. Re:You could roll your own. by Vlado · · Score: 1

      Anytime you do a write, the entire stripe has to be available to recompute parity. If you aren't doing full stripe writes, that will often mean having to read data in from a portion of the drives.

      In RAID3 the whole stripe needs to be available for parity recalculation. That's why RAID3 is usually considered a bad solution since most of the applications have small, random I/O requirements. RAID3 is nice for large file operations where you would write the whole stripe at once anyway.

      RAID5 is a type of RAID that is also referred to as an "independent access" RAID. One of the benefits is actually the ability to only engage two physical drives for every write transaction: the drive where the data was changed and the drive where the parity resides. Since the parity is evenly distributed across all of the drives it means that it does not become the bottleneck as it would if you had RAID4, where the parity is on a dedicated drive. That's also the reason why RAID4 is usually not used anymore in modern storage systems.

      So the sum of this is that if you have a RAID5 setup that has few disks write performance will be bad relative to the RAID5 where you have a lot of disks.

  3. Cmon people... by Creepy+Crawler · · Score: 5, Informative

    You might as well build it yourself.

    Go get a lowbie Core2, mobo, good amount of ram, and 4 1TB disks. Install Ubuntu on them with LVM and encryption. Run the hardening packages, install Samba, install NFS, and install Webmin.

    You now have a 100% controlled NAS that you built. You can also duplicate it and use DRBD, which I can guarantee that NO SOHO hardware comes near. You also can put WINE on there and Ming on your windows machines for remote-Windows programs... The ideas are endless.

    --
    1. Re:Cmon people... by Anonymous Coward · · Score: 0

      Mac users aren't helpless babies. I've been a Mac user since the early 90s and I've built plenty of PCs. It's not exactly rocket science, you know.

    2. Re:Cmon people... by emmons · · Score: 1

      I'll second this with a couple notes:

      Encryption isn't so important unless you're worried about someone coming in and physically stealing your hardware, but it will complicate setup a bit and will slow down IO a bit (depending on CPU speed).

      Webmin is great for this type of thing.

      Your network connection is the limiting factor here. On large sequential reads, modern SATA drives with a mobo's onboard controller can easily maintain the 100MB/s or so it takes to max out your gigE connection.

      Spend your money on some decent networking kit that can actually get you close a full 1 Gb/s. If you're only getting 6 MB/s with your current 100Mb/s hardware that tells me you've got some crappy hardware and/or bad wiring. Work on that.

      --
      Do you even know anything about perl? -- AC Replying to Tom Christiansen post.
    3. Re:Cmon people... by SanityInAnarchy · · Score: 4, Insightful

      If they have a windows machine for "compatability testing" and the rest of the units are Macs, you know damn well this guy couldn't "build his own"!

      For what it's worth, I have worked in a place that almost exactly matches that description -- ton of macs, some leftover Windows PCs (rarely if ever used), and I ran Linux.

      Everyone in that office could have built their own, if they had a reason to.

      It is possible to actually like a Mac and not be technically illiterate / incapable of assembling a PC.

      --
      Don't thank God, thank a doctor!
    4. Re:Cmon people... by Anonymous Coward · · Score: 2, Informative

      Yeah, it's not like a mac runs UNIX or has a freeBSD userland with a full ports tree or anything.

    5. Re:Cmon people... by Anonymous Coward · · Score: 0

      If he's looking for maximum throughput, I would recommend against encrypting the entire volume unless there is some pressing need to do so.

    6. Re:Cmon people... by Creepy+Crawler · · Score: 1

      ---Encryption isn't so important unless you're worried about someone coming in and physically stealing your hardware, but it will complicate setup a bit and will slow down IO a bit (depending on CPU speed).

      Yeah, it is a hit on I/O, but if we're using a Core2Duo, there's a bit of CPU available.. And you can sell it as "All your data is encrypted on the disk as per Sarbanes Oxley/HIPPA/governmental org standard." It's not terribly that important, but a selling point.

      ---Webmin is great for this type of thing.

      Very true. Webmin is invaluable for this kind of stuff. It's easy to train somebody else what you're doing and basic checkups and stuff.

      ---Your network connection is the limiting factor here. On large sequential reads, modern SATA drives with a mobo's onboard controller can easily maintain the 100MB/s or so it takes to max out your gigE connection.

      Of course, it does depend on the network I/O, but that would depend on the types of users (s)he is serving. If they're just document users, it will be plenty of random reads and writes, and might call for 2 128GB SSDs for main disk, while offloading bigger files to the spindles. Still, considering a good managed switch with 2 dual homed 1Gbps might be called for. It comes down to a tradeoff of cost vs performance.

      --
    7. Re:Cmon people... by Creepy+Crawler · · Score: 1

      Yeah, it is a hit on I/O, but if we're using a Core2Duo, there's a bit of CPU available.. And you can sell it as "All your data is encrypted on the disk as per Sarbanes Oxley/HIPPA/governmental org standard." It's not terribly that important, but a selling point.

      Many businesses like to take lip service of "security", so feed it to them. Honestly, encryption would prevent gaining access to data if physically stolen, so would only make sense on laptops, not servers in a locked room. But it still looks good after the fact of a data breach.

      --
    8. Re:Cmon people... by swillden · · Score: 5, Informative

      Your network connection is the limiting factor here. On large sequential reads, modern SATA drives with a mobo's onboard controller can easily maintain the 100MB/s or so it takes to max out your gigE connection.

      I second this.

      A good way to test your network connection is with netcat and pv. Both are packaged by all major Linux distos.

      On one machine run "nc -ulp 5000 > /dev/null". This sets up a UDP listener on port 5000 and directs anything that is sent to it to /dev/null. Use UDP for this to avoid the overhead of TCP.

      On the other machine, run "pv < /dev/zero | nc -ulistenerhost 5000", where "listenerhost" is the hostname or IP address of the listening machine. That will fire an unending stream of zero-filled packets across the network to the listener, and pv will print out an ongoing report on the speed at which the zeros are flowing.

      Let it run for a while and watch the performance. If the numbers you're getting aren't over 100 MB/s -- and they often won't be, on a typical Gig-E network -- then don't worry about disk performance until you get that issue fixed. The theoretical limit on a Gig-E network is around 119 MBps.

      Do the same thing without the "-u" options to test TCP performance. It'll be lower, but should still be knocking on 100 MBps. To get it closer to the UDP performance, you may want to look into turning on jumbo frames.

      pv is also highly useful for testing disk performance, if you're building your own NAS (highly recommmended -- a Linux box with 3-4 10K RPM SATA drives configured as software RAID0 array will generally kick the ass of anything other than very high end stuff. It's nearly always better than hardware RAID0, too).

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    9. Re:Cmon people... by swillden · · Score: 4, Informative

      pv < /dev/zero | nc -ulistenerhost 5000

      Slashdot at the space after "-u". That should be "pv < /dev/zero | nc -u listenerhost 5000".

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    10. Re:Cmon people... by Anonymous Coward · · Score: 0

      Possible, but definitely not "likely" in my experience.

      Disclaimer: I'm a Windows, Linux, and Mac PC owner with 10 years network admin and technician experience.

    11. Re:Cmon people... by ceoyoyo · · Score: 1

      I used to make a decent amount of money running a small company that mainly built custom PCs for people. At home I have entirely Macs now.

    12. Re:Cmon people... by gad_zuki! · · Score: 1

      Building it yourself is the optimal solution. We just bought a Netgear ReadyNas. Its maxes out at 25/MBs. Thats not a network limition, thats the raid card/CPU limitation. The little beater CPU in there doesnt come close to the cheapest C2D.

      Smallnetbuilder.com has tons of reviews and rarely do these things break 30/MBs. If speed is the most important thing then you gotta build your own. Grain of salt of course, the guy who reviews this stuff there is quick to sing its praises and not piss off the companies that advertise on his site and give him free stuff. Avoid Buffalo like the capacitor plague.

      Of course there are advantages with a turn-key solution. I spent very little time with the readynas and it does everything we need. Cute little rackmount too.

    13. Re:Cmon people... by rthille · · Score: 2, Funny

      slashdot apparently also ate the 'e' in 'ate' :-)

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    14. Re:Cmon people... by rthille · · Score: 1
      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    15. Re:Cmon people... by xthor · · Score: 2, Informative

      A good way to test your network connection is with netcat and pv. Both are packaged by all major Linux distos.

      Another network speed test that is cool, although not included with a distro, is netio. It's cross-platform, with versions for Linux and for Windows in the archive. I just wish I could find a version for my MacBook (or, knew enough to get it to compile under fink).

    16. Re:Cmon people... by TCM · · Score: 1

      Why is everyone always choosing "odd" numbers of disks like 4 or 8 for RAID5?

      The optimum number of components for a RAID5 is 2^n+1, i.e. 3, 5 or 9, although personally, I would stop at 5.

      With 2^n data components, you can match filesystem block sizes perfectly to the stripe size, avoiding unnecessary reads on writes.

      --
      Of course it runs NetBSD. BTC: 1NT7QvbetmANwaMzhpVL6
    17. Re:Cmon people... by dbug78 · · Score: 1

      You also can put WINE on there and Ming on your windows machines for remote-Windows programs... The ideas are endless.

      Could you explain this? The only "ming" I'm able to find is a SWF creation library.

    18. Re:Cmon people... by Anonymous Coward · · Score: 0

      Don't flatter yourself!

    19. Re:Cmon people... by Creepy+Crawler · · Score: 1

      I meant XMing..

      The free MS Windows X server.

      --
    20. Re:Cmon people... by Lumpy · · Score: 1

      Exactly. I love MACS.. and I know for a fact I can run circles around the GP poster in computer building. I have built my own Pentium III class X86 PC from the chip level and a Alpha Based embedded computer. (I miss the resources of big college) I designed my own motherboards. I know EE's that design Motherboard Chipsets that love MACS.

      In fact most MAC users and lovers I know personally are far more knowledgeable about IT, PC's and MAC, as well as EE than most windows users I know. But then I hang with EE people.

      --
      Do not look at laser with remaining good eye.
    21. Re:Cmon people... by SanityInAnarchy · · Score: 1

      Possible, but definitely not "likely" in my experience.

      Perhaps not... Let me count.

      I know a good five or so Mac users who refuse to consider anything except a Mac... because it's what they're used to. Because they are trained by rote, and seriously, if you were to move something on their desktop, they'd be confused -- or rather, that would have been the case a few months ago. Now, EVERYTHING is on their desktop, several layers deep, and they can't find anything.

      I also know a good five or so Mac users who run Windows in a virtual machine, know about Spaces and use it, develop software for a living, build things out of wires and microcontrollers in their spare time (think Arduino), and certainly are not afraid of opening a PC and digging through it. One has a Mac Pro at home.

      I'm sure they could all give you better reasons than I about why they like Macs -- I mostly run Linux, mostly on PC hardware. Maybe it's just the people I hang out with, but I seem to get about equal numbers of Mactards and Mac gurus.

      --
      Don't thank God, thank a doctor!
    22. Re:Cmon people... by naasking · · Score: 1

      Rolling your own doesn't get you all the automation that come with packaged solutions for free. Stuff like all the supported protocols (SMB, AFP, HTTP, Rsync, etc.), automatic mirroring, pluggable UPS, and expandable RAID disks (see the ReadyNAS Pro -- seems like a decent SOHO solution).

    23. Re:Cmon people... by Anonymous Coward · · Score: 0

      > If they're just document users, it will be plenty of random reads and writes, and might call for 2 128GB SSDs for main disk, while offloading bigger files to the spindles.

      Read the post.. he's asking for a server for moving around small numbers of large files.

      That's not to say that SSDs aren't sweet as hell though if you can afford them. :)

    24. Re:Cmon people... by SoupIsGoodFood_42 · · Score: 1

      Yeah, so much less hassle than buying a single unit that does it all for you.

      Perhaps this person has other things to do?

    25. Re:Cmon people... by swillden · · Score: 1

      LOL. Apparently.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    26. Re:Cmon people... by Anonymous Coward · · Score: 0

      ...or try iperf -- much easier, more repeatable.

      http://iperf.sourceforge.net/

    27. Re:Cmon people... by swillden · · Score: 1

      Looks like it should be a nice tool, but at least with the default options it reports a much lower bandwidth on my network. nc can move 98 MBps in TCP but iperf only gets 52 MBps. In UDP mode, nc can move 113 MBps, but iperf gets an abysmal 0.13 MBps. That's right, 130 KBps.

      Very strange.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    28. Re:Cmon people... by Anonymous Coward · · Score: 0

      Looks like it at the as well...

    29. Re:Cmon people... by Sentry21 · · Score: 1

      I'd love to know why you would suggest using encryption on drives that are going to be shared over the network.

      Installing OpenFiler does everything you've suggested, with the exception of installing Webmin, which is garbage and for which OpenFiler has a much better replacement geared specifically towards managing network shares, LVM mounts, etc.

    30. Re:Cmon people... by AngelofDeath-02 · · Score: 1

      This has some really nice information - I saved it.

      My experience shows a default transfer rate of:
      tcp 73.5MB/s
      udp 88.8MB/s

      When enabling jumbo frames on both systems:
      tcp 103MB/s
      udp 115MB/s

      However, I also noticed that this negatively impacted SMB transfer rates. They went from 24MB/s to about 9MB/s - so I suspect there's some optimizing to be done with samba as well. I'm not sure I like adjusting this as it might negatively affect systems that don't use jumbo frames... Do you have any experience configuring samba to work well with this?

      --
      No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
    31. Re:Cmon people... by AngelofDeath-02 · · Score: 1

      I should add that I have previously experimented with TCP_NODELAY SORCVBUF=8192 SO_SNDBUF=8192 and this also negatively impacted file transfers in a similar way.

      --
      No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
    32. Re:Cmon people... by Wiseleo · · Score: 1

      Minor thing... Please avoid the Seagate 1TB drives for the time being. You'll thank me for this.

      --
      Leonid S. Knyshov
      Find me on Quora :)
    33. Re:Cmon people... by hearnz · · Score: 1

      I bought one of these a few months ago, and stuck 5 x 1Tb drives in it. It's certainly not the cheapest solution, but I don't regret it for a minute - it's fast (I get pretty similar speeds to those reported in the SmallNetBuilder review), quiet (can just barely hear the fan, and there is very little drive noise), reliable, easy to set up and use, and the fact that I can simply SSH into it and install/run whatever I like on there is a huge bonus. It has dual Gbit NICs and supports Samba, NFS and iSCSI, and a whole truckload of other features. I highly recommend it to anyone looking for something like this.

    34. Re:Cmon people... by Creepy+Crawler · · Score: 1

      OOOOh.. Sounds like a great idea. NOT!

      Openfiler: The project that charges 30 Euro for the fucking non-printing PDF administrator handbook.

      --
    35. Re:Cmon people... by sproot · · Score: 1

      most .. lovers I know personally

      I'd hope you know all your lovers personally!

      Oh wait

    36. Re:Cmon people... by thornomad · · Score: 1

      Thanks for that tip for measuring network performance speed, am interested to check it out tonight on my home system...

      I'm not a sysadmin -- I do run a NAS on Ubuntu server from the basement at home (using netatalk/AFP -- I have all Mac OS X machines upstairs) ... but I still haven't found a straightfoward way to measure my NAS's file sharing performance from the command line ... I have seen things like Bonnie++ (which I find too complicated) or just using "time cp" ... but is there an easier way to get some straightforward performance data from my NAS ?

      I haven't found it yet, in my Goooogle-ing. Everyone here is saying they are seeing this or that XXX/mbs ... but what I don't understand is how they measure this ... unless they are using a stopwatch or something.

      Thanks for your tips.

    37. Re:Cmon people... by rthille · · Score: 1

      Oh good, a person with real experience...
      Do you have any OSX clients? If so, does the AFP implementation/version support the latest version of the protocol? (Long filenames, etc) I tried to find the AFP version on the specs, but it's not listed, and the forum where you can ask pre-sales questions never mailed me my signup info. I assume because of the graylisting my ISP does.

      Also, have you tried out any of the email packages? I'm interested in replacing my old NetBSD/Cobalt MIPS server (slow as hell), and so I'd want to install an email server (SMTP & IMAP at least), TMDA, etc.

      Also, I wonder if I could move the http server to another port, or just run it on one of the interfaces and run Zope on the other one for external serving.

      Basically, would I be happy with it as a general purpose Linux box, that just happens to have a really nice form factor for my purposes :-)

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    38. Re:Cmon people... by pjacob · · Score: 1

      Great post. Saved the tip for future reference.

    39. Re:Cmon people... by swillden · · Score: 1

      The "pv" tool is a great way for doing all sorts of ad-hoc throughput measurements. To measure read performance, just do something like "pv /dev/null". Assuming filename is on the NAS and isn't cached, you'll get a good measurement of read performance. To test write performance, just pull from /dev/zero (or /dev/urandom, to avoid any false efficiency that might be gained from compression -- be sure to test /dev/urandom performance first, though) and write to the NAS: "pv filename".

      I use pv all the time, for all sorts of things. Mainly because I'm always curious about how fast stuff is going, but also because it's often useful with long tasks to get an idea about how long it will take. Just this morning I had to untar a 6 GB tar file, so I ran "pv file.tar.bz2 | tar xvf -".

      However, for file system performance, I'd probably use bonnie++. It's not really that complicated. There are a lot of options, but you can just run it without any options at all and get useful information out. For determining maximum throughput, look at the sequential read and write numbers. The per character numbers aren't very useful for most people, so you might want to add the "-f" (for "fast") option, to skip those time-consuming and not-very-useful tests.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    40. Re:Cmon people... by hearnz · · Score: 1

      I don't have any OSX clients, sorry, or much experience with OSX in general. But the NAS itself is basically just a Linux box (it runs BusyBox), and uses ext3 filesystems and Samba/NFS. So I presume that OSX clients would work just as well (or badly, if that is the case) as they would with any similar Linux setup.

      I don't run a great deal of the out-of-the-box software on mine, as I already have a couple of Linux servers for those sorts of duties. I mainly use mine for backups, music/video storage for an HTPC, and iSCSI targets for a bunch of VMware machines. But you can run pretty much whatever you want on it - there's a package management system built in which makes it easy to install a whole truckload of third-party prepackaged software (check out http://ipkg.nslu2-linux.org/feeds/optware/ts509/cross/stable/), or you can just use that to install gcc and then compile anything else yourself.

    41. Re:Cmon people... by thornomad · · Score: 1

      Thanks for that - I will check out the pv tool a little more carefully; also, I won't give up on Bonnie yet ... I guess I was a little overwhelmed by the options and, since I am just at home, I didn't have a lot to compare it to. If I upgrade my NAS server (which is running on an old 500mhz Dell) I do want to get a lot of measurements before and after the upgrade ... to see what I've gained.

      I won't give up yet! And, thanks.

    42. Re:Cmon people... by rthille · · Score: 1

      I'd go to 10 drives with RAID6 and a good scheme for testing that all the blocks are good, and a remote backup.

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    43. Re:Cmon people... by Zantetsuken · · Score: 1

      for what reason? I'd actually bought 2x of those a while back and stuck them in a D-Link DNS323 (2 bay soho NAS)...

      The only problem I'd seen on them is tons of people with nVidia raid controllers it won't work, but after they saw the reviews saying it doesn't and they still buy it, that's their fault...

    44. Re:Cmon people... by esarjeant · · Score: 1

      Or you could just run bing:

      http://fgouget.free.fr/bing/bing_src-readme.shtml

      This is a point-to-point bandwidth measuring tool based on ICMP. It works fairly well and doesn't require you to saturate the network to establish thruput.

      It's also nice because you can use bing to measure the speed between two arbitrary hosts. In other words, Server A and Desktop A or Server A and Desktop B.

      With that said, if you do observe a potential issue based on your bing results then you will want to try a method like Shawn suggests here (ie; saturate the pipe and see what results from that).

      As an aside, if you can use iSCSI or NFS for your network file solution the performance over something like CIFS is notable -- especially on slower networks.

      --

      Eric Sarjeant
      eric[@]sarjeant.com

    45. Re:Cmon people... by ZerdZerd · · Score: 1

      I suggest iperf.

      iperf -s on the server side
      iperf -c <server ip> on the client side.

      --
      I'm not insane! My mother had me tested.
    46. Re:Cmon people... by jon3k · · Score: 1

      I always recommend using iperf, which you can find in the repos for most modern linux distros (ubuntu, fedora, etc).

      On the server, just $iperf -s

      On the client: $iperf -c [ip of server]

    47. Re:Cmon people... by jon3k · · Score: 1

      Just do RAID1 please. RAID0 really isn't worth it. http://www.anandtech.com/storage/showdoc.aspx?i=2101

    48. Re:Cmon people... by Just+Some+Guy · · Score: 1

      In fact most MAC users and lovers I know personally are far more knowledgeable about IT, PC's and MAC, as well as EE than most windows users I know. But then I hang with EE people.

      Can they spell "Mac"?

      --
      Dewey, what part of this looks like authorities should be involved?
    49. Re:Cmon people... by swillden · · Score: 1

      For some reason, iperf shows much less throughput on my network than nc. Half for TCP and 1/1000th for UDP mode. I wonder why that is.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    50. Re:Cmon people... by swillden · · Score: 1

      Just do RAID1 please. RAID0 really isn't worth it.

      I have to disagree, for three reasons.

      First, write performance of RAID0 is inevitably much higher than RAID1. You don't see this as much with hardware RAID cards (which is what was used in the article you referenced), but with Linux software RAID, the difference is just what you'd expect: If you have N disks that each handle M MBps, RAID0 gives you almost N*M and RAID1 gives you a little less than M.

      Second, read performance of RAID0 is also significantly higher than with RAID1, even though intuition would predict they should be about the same. On every system I've tested, RAID0 read performance is nearly double RAID1 read performance.

      Third, storage. Particularly if you're looking to get 300 or 400 MBps read rates (yes, I know you can't stuff that much down a Gigabit pipe, but it's still useful in many cases, and you want to have more disk bandwidth than network bandwidth to make sure you can keep the network full even when the disks are doing something else), mirroring three, four or five drives is very expensive in terms of the cost per GB.

      Of course, a four or five-disk RAID0 puts you at high risk of a failure, which will cause you to lose everything. So if you're doing that, you need to either go with RAID10, which will only double the number of disks you need, or else have a good backup system in place.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    51. Re:Cmon people... by jon3k · · Score: 1

      I don't know, that's pretty strange. I can easily saturate fastethernet and 1GbE using it.

    52. Re:Cmon people... by jon3k · · Score: 1

      As far as RAID0 vs RAID1 read performance, it depends on whether or not the RAID controller will do striped reads on the RAID1 array. If it will perform striped RAID1 reads (which I don't believe md will), the performance will be nearly identical. Also, in practice, the controller will actually split requests between disks, so while single threaded benchmarks look bad, real world performance is actually much better.

      As far as having a backup: RAID is not backup. RAID doesn't help you when the filesystem corrupts itself or you accidentally delete something. They aren't alternatives for one another. RAID is for high availability and performance and backups are for recovery.

    53. Re:Cmon people... by swillden · · Score: 1

      As far as RAID0 vs RAID1 read performance, it depends on whether or not the RAID controller will do striped reads on the RAID1 array. If it will perform striped RAID1 reads (which I don't believe md will)

      According to the Linux Software RAID FAQ, it does. More precisely, the FAQ says it performs "balanced" reads. And it does provide better read performance in practice than a single-disk, even on single-threaded benchmarks, but not nearly as good as RAID0.

      As for whether or not it performs better in real-world usage than in benchmarks -- that depends on the workload. The workload described in the original question sounds like one that will be doing a lot of large, sequential single-threaded reads, which is exactly what benchmark tools usually do.

      And, of course, there are still the issues of write performance and storage cost.

      For example, a colleague of mine recently set up an OLAP server which must be able to store over 1 TB of data, all of which can be lost without issue, because it's just an extract from another system. He hoped to get 500 MBps sequential read throughput for large table scans, and good random access read performance as well.

      The solution was six 300 GB 15K RPM SCSI-320 drives, three on each of two SCSI controllers, with Linux software RAID0 (could also have been done with HW RAID, but testing showed that to be slower, both on benchmarks and on the real workload). How would you do that with RAID1? Six 1.5 TB drives? In practice you'd use RAID10, of course, with 12 of the drives. But only 6 drives fit in the 2U system, so you'd need to add an extra set of drive bays, with cabling, cooling, PSU, etc. The cost of that, plus the cost of the six extra drives adds $3-4K to the price of the system. And for what? To avoid a few hours of downtime while the system is reloaded?

      There are cases where RAID0 makes perfect sense. A server that stores static test data sets is another, particularly if it can be restored from backup in a timely fashion.

      As far as having a backup: RAID is not backup.

      Of course not. But if you're using RAID0, it's important to have a system in place that backs up very frequently, because your odds of a failure are high. The probability of filesystem corruption (assuming a decent file system) is low enough that you can afford a little less frequency. As for wanting back stuff that you deleted, that's a service level issue to be negotiated between the admin and the users. In many environments it's a non-issue with respect to backup frequency, because the policy is that if you delete stuff you wanted to keep, you're probably screwed.

      That said, for my home use, I have no practical way to back up the amount of data on my system so I rely primarily on RAID6 for reliability. It's an approach that's worked well for over 10 years now. I do have to be careful not to delete stuff I want to keep, of course, and for the subset of the data that is most critical, I keep copies on multiple machines. For a business environment, I agree that backups are essential.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    54. Re:Cmon people... by amRadioHed · · Score: 1

      I love MACs too, especially when it's late at night and I need some cash... oh you meant Macs. Never mind.

      I'm curious where this MAC thing comes from. Mac is short for Macintosh, it's never been an acronym.

      --
      We hope your rules and wisdom choke you / Now we are one in everlasting peace
    55. Re:Cmon people... by Anonymous Coward · · Score: 0

      Also - Get some good NICs and bond them together for more network throughput.

    56. Re:Cmon people... by jon3k · · Score: 1

      "According to the Linux Software RAID FAQ, it does. More precisely, the FAQ says it performs "balanced" reads. "

      Balanced read != striped read.

      Balanced read means read request 1 goes to disk 1 read request 2 goes to disk 2. If read request 1 was 10x the size of read request 2 it's not very useful. Striped read means if you ask for read request 1, that single request is split between two disks. There is no comparison between the performance for single threaded reads (obviously).

      "The solution was six 300 GB 15K RPM SCSI-320 drives, [...]"

      I really have no idea how this has anything to do with me recommending RAID1 vs RAID0 for his HOME NAS DEVICE.

      "Of course not. "

      Ok well your original post said RAID10 _OR_ have a good backup system in place.

      "In many environments it's a non-issue with respect to backup frequency, because the policy is that if you delete stuff you wanted to keep, you're probably screwed. "

      That's insane. So if User A deletes incredibly important company information, well, it's his fault. Great, fire him. Now what?

      I will however agree with you on RAID6. You take a little performance and capacity hit over RAID5 but it's worth it.

    57. Re:Cmon people... by swillden · · Score: 1

      Balanced read means read request 1 goes to disk 1 read request 2 goes to disk 2. If read request 1 was 10x the size of read request 2 it's not very useful. Striped read means if you ask for read request 1, that single request is split between two disks. There is no comparison between the performance for single threaded reads (obviously).

      And yet MD RAID1 does have better performance on single-threaded reads than single-disk reads. I think there's a bit more here than what you're describing.

      "The solution was six 300 GB 15K RPM SCSI-320 drives, [...]"

      really have no idea how this has anything to do with me recommending RAID1 vs RAID0 for his HOME NAS DEVICE.

      First, it wasn't a home NAS device, it was a small office NAS device.

      Second, the same principles apply, just scaled down. For a given capacity and performance requirement RAID1 will cost significantly more than RAID0. Particularly if the capacity required is more than the size of commonly-available drives.

      "Of course not. "

      Ok well your original post said RAID10 _OR_ have a good backup system in place.

      Which I explained in the previous post. You seem to be assuming that the alternative to a good backup system is no backup system? That's a reasonable misunderstanding, but I clarified that by "good" I was referring to backup frequency. Poor wording on my part.

      "In many environments it's a non-issue with respect to backup frequency, because the policy is that if you delete stuff you wanted to keep, you're probably screwed. "

      That's insane. So if User A deletes incredibly important company information, well, it's his fault. Great, fire him. Now what?

      It's a cost/benefit tradeoff, and many companies do prefer to put the onus on the employee not to delete stuff that he needs. You can call it insane if you like, but it's very common in the industry, even among the rare companies who have competent IT departments. Obviously, if the data in question is "incredibly important", then the cost/benefit analysis says backups should be implemented.

      I will however agree with you on RAID6. You take a little performance and capacity hit over RAID5 but it's worth it.

      Absolutely. Particularly since the process of resyncing a degraded array touches every sector of every disk and is highly likely to uncover any problems with any of the remaining drives -- and if a second drive goes down in a RAID5 array, you're hosed.

      I speak from personal experience here. My home server (which, as you may recall, is not backed up) used to be RAID5. One of the disks died hard one day, and the system started rebuilding onto the hot spare I had recently added. During the rebuild a transient failure occurred on another drive. Boom. Data All Gone. Well, not exactly.

      I turned the machine off and mulled it over for a week, then went back to the e-mails mdadm had sent me, which contained the order of the drives in the array when it died. I forcibly reassembled the array, telling MD to assume it was clean (though degraded). It restarted the rebuild process, but failed again. On the third try, however (yes, I was desperate), the rebuild succeeded. I didn't lose a single byte, as far as I can tell.

      Based on that experience, I switched from RAID5 with a hot spare to RAID6, and implemented periodic array checking (MD checkarray basically simulates a resync). I actually have a mixed system now, some RAID0 for high-performance scratch space, some RAID5 for semi-important but replaceable stuff and some RAID6 for important stuff. A subset of the RAID6 data is regularly synced to other machines in the house. If the house burns down, though, I'm screwed unless I can grab my laptop on the way out, and I'm going to be much more focused on making sure the kids get out. I'm trying to get a family member to let me sync stuff to their machine for an off-site backup.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    58. Re:Cmon people... by jon3k · · Score: 1

      "And yet MD RAID1 does have better performance on single-threaded reads than single-disk reads. I think there's a bit more here than what you're describing."

      That would be interesting to see, do you have any benchmarks? That doesn't jive with my (limited) understanding of how MD RAID1 handles read requests.

      "Second, the same principles apply, just scaled down."

      Again, a several disk raid5 solution really doesn't apply here.

      "It's a cost/benefit tradeoff, and many companies do prefer to put the onus on the employee not to delete stuff that he needs"

      Many companies also don't implement firewalls or don't regularly patch systems. That doesn't make it a recommended business practice. _Most_ companies take daily incremental/differential backups along with good, regular (usually weekly) full back ups. As is recommended by any competent IT professional.

      "Which I explained in the previous post. You seem to be assuming that the alternative to a good backup system is no backup system? That's a reasonable misunderstanding, but I clarified that by "good" I was referring to backup frequency. Poor wording on my part."

      So I should assume your other suggestion was to use RAID10 and a "bad backup solution" ? Of course the alternative to a "good backup solution" would be no backup solution. Anyway, semantics, getting nowhere.

      Yes, RAID6 is fantastic, RAID-Z is also pretty amazing and a little company called lime-technology makes a great piece of software called unRAID. It lets you mix and match disks, uses a system similar to RAID-4 but each disk has it's own filesystem. So you could lose every disk but one and the data on that disk is still readable. Great for a small home setup for media and such.

      http://lime-technology.com/

    59. Re:Cmon people... by swillden · · Score: 1

      "And yet MD RAID1 does have better performance on single-threaded reads than single-disk reads. I think there's a bit more here than what you're describing."

      That would be interesting to see, do you have any benchmarks? That doesn't jive with my (limited) understanding of how MD RAID1 handles read requests.

      All I have is the numbers from my home server, using bonnie++. RAID1 over four disks (all similar, though not identical) is about 130% of the speed of a single disk, and RAID0 gives me 250% of single-disk speed.

      "Second, the same principles apply, just scaled down."

      Again, a several disk raid5 solution really doesn't apply here.

      RAID0, not RAID5. The system I mentioned was six-disk MD RAID0. RAID5 would have been slower, and the redundancy pointless.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    60. Re:Cmon people... by SanityInAnarchy · · Score: 1

      Probably some confusion about MAC addresses. And it is irritating...

      --
      Don't thank God, thank a doctor!
  4. None by afidel · · Score: 4, Interesting

    If you want decent throughput build it yourself. Seriously. I have a coworker that bought 5 different NAS devices to do a bakeoff for a small skunkworks office and they all sucked for throughput. We ended up buying a $1K NAS that still wasn't great but sure beat all the SOHO ones. Numbers were ~8MB/s max on the fastest SOHO unit vs 25MB/s on the midrange one.

    --
    There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
    1. Re:None by ColdWetDog · · Score: 1

      That mirrors my experience with a ReadyNAS NV+ - reliable, not particularly cheap, slow. For my purposes (just a backup of terabytes of photographic images) it's fine. For anything needing throughput, I'd roll my own.

      I'd avoid Drobo as well. Although cute and brainless it's really not a NAS (has to be hooked to firewire or BogForbid, USB2. Software is proprietary and they use a non standard RAID format.

      --
      Faster! Faster! Faster would be better!
    2. Re:None by SanityInAnarchy · · Score: 1

      Software is proprietary and they use a non standard RAID format.

      Sounds exactly like ReadyNAS.

      Not that it matters, as a user -- doesn't it just present itself as a mass storage device, no software needed on the host box?

      --
      Don't thank God, thank a doctor!
    3. Re:None by Capt.DrumkenBum · · Score: 1

      Am I the only one who read this as "just a backup of terabytes of pornographic images"?

      --
      If I were God, wouldn't I protect my churches from acts of me?
    4. Re:None by _pi-away · · Score: 1

      Drobo does have a NAS attachment, but the performance will be poor, especially in a gigE content.

      --

      "The crows seemed to be calling his name, thought Caw."
    5. Re:None by _pi-away · · Score: 1

      context, rather.

      --

      "The crows seemed to be calling his name, thought Caw."
    6. Re:None by tacocat · · Score: 1

      Search for Jack Rabbit media storage. It's relatively cheap but can easily manage GigE speeds. It's a small company a friend of mine has going and is capable of exceeding EMC brochures. When I say relatively cheap that's against the big boys, not SoHo.

      But it's definitely a kickin' box.

    7. Re:None by styrotech · · Score: 1

      Sounds exactly like ReadyNAS.

      Only slighty. You can configure (through the standard UI even) a ReadyNAS to use standard Linux software raid (and ext filesystems) rather than their proprietary xraid stuff if you care about that kind of thing. It is still supported and the disks are then portable to another Linux machine.

      And they run Debian Sarge - you can ssh in and apt-get all kinds of other stuff onto the box.

      Hardly "exactly" like a Drobo.

      The main problem with the ReadyNAS (and most similar units) is they are very expensive for the performance level (or lack thereof) you get. rsyncing large amounts of data just crawls.

    8. Re:None by Anonymous Coward · · Score: 0

      The NAS version of the Drobo is a little plugin box called Droboshare. It has a development SDK (recently released), and one of the first apps was, obviously, sshd. So now you can ssh into it and hack around as much as you want.

    9. Re:None by the_crowbar · · Score: 1

      The problem with these NAS boxes is the processor is only able to push about 25mb/sec. The whole point of the rsync protocol is to trade network bandwidth for CPU. If your CPU is already taxed then what do you think rsync is going to do? Rsync will crawl on any of these SOHO NAS boxes.

      --
      Have you read the Moderator Guidelines
    10. Re:None by ColdWetDog · · Score: 1

      Am I the only one who read this as "just a backup of terabytes of pornographic images"?

      I hope so. Perhaps a bigger monitor or better glasses are in store for you sir?

      --
      Faster! Faster! Faster would be better!
  5. dedicated PC by spire3661 · · Score: 4, Informative

    In terms of cost/benefit ratio, nothing beats a stripped down PC with a lot of drives stuffed in it or in an external esata enclosure. I run a HP NAS MV2020, and a linksys NAS200 and they both cant hold a candle to a PC in throughput. Ive heard of some commercial systems out there, but they cost a small fortune. Just my $.02

    --
    Good-bye
    1. Re:dedicated PC by JCSoRocks · · Score: 1

      I've had a similar experience. When I decided I wanted to make myself a small home file server I just took an old 3GHz P4 and put it into a new case with a big hard drive cage. Add a SATA card and a couple terabyte drives and I've got a nice Gigabit NAS setup with a much faster processor / RAM than anything you're going to get in a consumer level NAS. Best of all, I only had to spend money on the hard drives and the SATA card. Everything was already laying around.

      --
      You are using English. Please learn the difference between loose and lose; they're, there, and their; your and you're.
    2. Re:dedicated PC by willow · · Score: 0

      Is this OK?

      "They're saying their sister is loose but you're going to lose if you go there."

      --
      Moderation in everything, including moderation.
    3. Re:dedicated PC by willow · · Score: 0

      Amend to:

      "They're saying their sister is loose but you're going to lose your head if you go there."

      --
      Moderation in everything, including moderation.
    4. Re:dedicated PC by amiga500 · · Score: 1

      Unless you consider the cost of your electricity. That old PC will probably suck down way more power over a year than any of the SoHo NAS boxes.

  6. HP Media Vault may work out for you by mandark1967 · · Score: 1

    According to the CNET reviews I saw,located at http://reviews.cnet.com/external-hard-drives/hp-media-vault-mv2020/4505-3190_7-32104518-2.html the actual 5GB copy tests they did show it being faster than the rest, even one other system using RAID5

    --
    Sig Follows: "Suppose you were an idiot. And suppose you were a member of Congress. But I repeat myself." -- Mark Twain
    1. Re:HP Media Vault may work out for you by mandark1967 · · Score: 1

      Replying to state that the unit they reviewed, the mv2020, is compatible with MAC, Linux, and Windows, as shown below:

              * Min Operating system Linux, Apple MacOS , Microsoft Windows XP, Microsoft Windows 2000 Professional SP4

      More from the review: ...the Media Vault is quite a speed demon at both reading and writing. In the CNET Labs' tests, it averaged 15 minutes, 12 seconds to write a 5GB folder of mixed file types from the PC to the Media Vault. It read the same folder back in 13 minutes, 25 seconds. None of the other NAS units we've tested beat the Media Vault's read or write times.

      --
      Sig Follows: "Suppose you were an idiot. And suppose you were a member of Congress. But I repeat myself." -- Mark Twain
    2. Re:HP Media Vault may work out for you by Glonoinha · · Score: 2, Insightful

      I hate to point this out, but 5G in 15 minutes is about 5 megabytes per second.

      GigE peak theoretical throughput is like 125MB/s.
      Consumer grade hard drives can average throughput in the 60MB/s range.

      If this is the fastest NAS solution they tested and CNET is thrilled with their blazing 5MB/s sustained throughput to the NAS - I don't want one.

      I'm going to have to suggest going with a cheapo 2.8GHz HyperThreaded P4 based 'server' w/ GigE, 1G of RAM and a few SATA drives on a RAID controller. Use whatever OS you're familiar with, set it up as shared space and get the bandwidth your application needs.

      --
      Glonoinha the MebiByte Slayer
    3. Re:HP Media Vault may work out for you by Anonymous Coward · · Score: 0

      I'm going to have to suggest going with a cheapo 2.8GHz HyperThreaded P4 based 'server' w/ GigE, 1G of RAM and a few SATA drives on a RAID controller. Use whatever OS you're familiar with, set it up as shared space and get the bandwidth your application needs.

      I somewhat disagree. I built a not-so-cheapo 3.4GHz P4 Gallatin-based server using a Supermicro P4SCT+ motherboard (Intel 875 chipset with CSA Intel GigE networking) and a 3ware 9550SX hardware RAID controller, and it can only manage to send at about 20% of GigE bandwidth. My newer Gateway laptop (2.2 GHz Core 2 Duo, PCI-e Marvell GigE networking) can manage to send at about 50-60% of GigE bandwith to the same server. Other systems that I have hooked up to the network (Acer 6930 laptop, HP a6660 desktop) behave similarly to the Gateway laptop.

      While I don't have a full explanation, it's clear that either one or both of a PCI-e based NIC and a dual core CPU are necessary to maximize throughput on a GigE link. Quad core is not necessary. Not having a socket 775 motherboard of any flavor (and no AMD equipment), I can't say whether the interface or CPU is more important, but I suspect that the CPU is more important since that server setup can at least receive at 50-60% of GigE bandwidth, and CSA has substanitally more bandwidth than PCI or PCI-X (266 MB/s essentially dedicated to the network interface).

    4. Re:HP Media Vault may work out for you by Barny · · Score: 1

      A Via EPIA based raid solution with a gig of ram and freeNAS loaded will generally max out your drives bandwidth, supports full drive encryption, a "slew" of shareing methods including iSCSI (host and target) NFS SMB iTalk FTP torrent.... yeah, you get the idea.

      And best of all, on these boards it chews very little power :)

      --
      ...
      /me sighs
    5. Re:HP Media Vault may work out for you by Glonoinha · · Score: 1

      That's very odd - I'm interested in which OS you were using, server configuration (ie, how much memory), what RAID level you were using, and with how many drives (and what drives you were using.) The benchmarks show that the 9550sx should be able to move data off the hard drives more than fast enough to saturate a GigE pipe.

      Sucking data off the hard drives and serving it up to a single client over the network isn't really a CPU intensive process. A 3.4GHz P4 shouldn't break a sweat doing this.

      --
      Glonoinha the MebiByte Slayer
    6. Re:HP Media Vault may work out for you by Glonoinha · · Score: 1

      Given the kind of hardware Novell Netware 4.x was designed to use and the level of performance it saw on that hardware - I'd be very interested in seeing what kind of performance we could get from it resurrected as a NAS server (assuming you could get drivers (NLM's) for current hardware, and a decent TCP/IP stack running on it.) I imagine Netware 4.02 or 4.11 would make a ~screaming~ NAS (pretty much that's all a Netware file server did, back in the day ...)

      --
      Glonoinha the MebiByte Slayer
    7. Re:HP Media Vault may work out for you by grub · · Score: 1


      I'll add a vote for FreeNAS, it rocks!

      I have a RAID5 running on 6x 750 GB drives (~3.3 TB usable space) feeding the house and, most importantly, our PopcornHour media device.

      --
      Trolling is a art,
  7. Drobo by rhuddusa · · Score: 0

    Just got a drobo myself with a droboshare front end. I did some testing and can get about 15 MB/s reads with 13 MB/s writes with 4 250 gig seagate hard drives. i was hoping for more speed, but traded it for the other features of the drobo. tested on a GigE network.

    1. Re:Drobo by Shawn+Parr · · Score: 1

      I also have a Drobo, and had a Drobo Share. In my situation the Drobo Share didn't make sense to keep.

      The Drobo v2 can transfer at over 30MB/s across its firewire. I didn't test the USB, but I imagine it is not too much less than that.

      The Drobo Share gets maybe up to 15MB/s, which is half of what the Drobo is capable of. If you want to use a Drobo, use a stripped down machine with a bit of RAM and a streamlined OS just for sharing. That way you can use the throughput of the Drobo more effectively.

      Of course at that point you could probably look at freenas and just drop sata drives into the machine you will have to use. If configured right you could probably get more than the 30MB/s the drobo can kick out.

      Of course the submitter wants an off the shelf solution, so unless he wants only 15MB/s, which is not very fast for even a small office load, the Drobo is probably out of the question for him.

      Of course with the drobo he would get data redundancy, and automounting capability on the macs and the PC with the dashboard software (if using the Drobo Share). But still there are better solutions for his needs.

  8. Some quick&dirty SMB file transfer numbers by Ekuryua · · Score: 1

    I have both a dlink DNS-323 and a readynas duo (netgear), and they're dealing fairly decently on GbE.(though the netgear does quite better) Typically that's ~30MB/s with my dns-323 in RAID1, and ~40MB/s with the netgear. Still far from full GbE saturation, but seeing how cheap they are, and easy to replace/manage(as well as being compact) they're quite useful imho.

  9. ReadyNAS by IceCreamGuy · · Score: 2, Informative

    We have a ReadyNAS 1100, it's alright, but I wouldn't call it stellar. I get around 80Mb/sec to it over the network, but the management interface is IE only (as far as I can tell, since it has problems with FF and Chrome), and it has these odd delays when opening shares and browsing directories. Some of the nice features are the out-of-the-box NFS support and small, 1U size.

    1. Re:ReadyNAS by Anonymous Coward · · Score: 0

      I manage my ReadyNas Duo entirely on a Mac, usually in Safari without issue. Are you on the latest RAIDiator?

      I'd say that mine keeps up with the network just fine. We have an Intel one at work, and it's as slow as a dog.

    2. Re:ReadyNAS by Anonymous Coward · · Score: 0

      I have 2 Infrant (Now netgear) ReadyNAS NV+ the throughput is 80MB/sec and the unit does have issues with Safari and Firefox 3.0 but the latest radiator update does fix that.

    3. Re:ReadyNAS by IceCreamGuy · · Score: 1

      You know I actually just assumed that the new RAIDiator would work the same with Webkit and didn't even try it out with Chrome since it was only a minor revision. Good to know.

    4. Re:ReadyNAS by ColdWetDog · · Score: 1

      I've been using the ReadyNas (NV+) web interface on FF 2 and 3 for a couple of months. Just checked, still works. I assume you mean the "Frontview" application?

      --
      Faster! Faster! Faster would be better!
    5. Re:ReadyNAS by fincan · · Score: 1

      For NV+ they released a new firmware which also fixes FF3 compatibility issues. Probably they did it for 1100 too. We were trying to build a backup server in the company I am working for and bought NV+ and ended up rolling our own linux server with Amanda (www.amanda.org). It is much faster than Netgear stuff and we used leftover computer parts, only new thing in it was 2x 750GB SATA disks in RAID 1 configuration (mirroring) and it has a write speed of ~50MB, which is limited by hard drive sustained write speed but it is more than enough for us.

    6. Re:ReadyNAS by Penguin's+Advocate · · Score: 1

      We have a couple of the 4TB ReadyNAS 1100's... 80MB/sec?!?!?!?!? How? What magic did you work to get that? The most we've been able to squeeze out of the things is ~30MB/s, which is in line with what most reviews have found. We've ended up just using them as on-line backups because they're too slow for active use. We've got them mounted via NFS to a linux box with 4TB of really fast RAID5 storage internally, and we just use rdiff-backup to sync the internal RAID daily. In terms of a speed comparison, the linux box (via an Intel PCIe quad gigE board) has no trouble pumping 350-400MB/s.

      --
      Frag 'em all...
    7. Re:ReadyNAS by verbalcontract · · Score: 1

      You may need to upgrade your firmware. I had problems with pre-4.00 firmware and Firefox 3. Now, we're running 4.01c1-p2 and the web interface runs fine.

  10. Worked on Buffalo's... by Anonymous Coward · · Score: 0

    I worked on an older version of Buffalo's NAS. It has a good, standard design (how you would probably do it if you were rolling your own anyway) and has good throughput, but it may try to do a lot of extra work if you are uploading media to it due to some media streaming capabilities in some versions.

  11. Dlink DNS-323 by speeDDemon+(nw) · · Score: 2, Insightful

    I have evaluated a few different products (I have a retail store) and so far I have been very happy with the DLINK DNS-323
    Disclaimer: I have no affiliation with DLINK other than I stock some of their goods

    1. Re:Dlink DNS-323 by Burdell · · Score: 1

      I had a DNS-323 and never could get what I would consider good throughput with it (why bother with gigabit when it can barely fill 100 megabit). I ended up building a cheap PC out of spare parts and a few new things for not a lot more than the DNS-323, and it performs much better.

    2. Re:Dlink DNS-323 by Anonymous Coward · · Score: 0

      Mod up. I can get 10MB/sec on 100mbit via on mine. And fun_plug is great.

    3. Re:Dlink DNS-323 by dropadrop · · Score: 1

      I had a DNS-313 for a week. Despite being on the official "OSX supported" list it would not work with OS-X 10.5. Was fine with Windows, Linux and OS-X 10.4, though speeds where nothing to write home about (around 10MB/s on a gigabit network).

      Really confirmed once again my opinion on dlink... I took it back and got a Qnap instead. Speeds are better and it has worked very well. The only reason I'd get a Qnap for a small company though would be the built in ability to use it as storage for IP video survalance cameras. I really think a proper server is the way to go if you actually require good speed from a cheap device.

    4. Re:Dlink DNS-323 by Radix999 · · Score: 1

      I have a DNS-323 and can attest that while it has a Gigabit ethernet card - the network speeds max out at about ~18-19MB/s read speed. This is when using two individual drives and no RAID.
      Accessing both drives at once causes the speeds to balance out at ~10MB/s on each drive.

      I've heard the Netgear ReadyNAS 2 drive one performs better - read a report from one guy on the OCAU forums recently that said he managed to get 39MB/s read and about 24MB/s write speed.

      --
      -- Wireless WaFreenet user since March 2002
  12. Not Buffalo by rworne · · Score: 1

    They have neat solutions, but their throughput is horrible. They support GigE, but the CPUs they use in their boxes are so underpowered they never achieve anything reasonably higher than 100-base-T (if that).

    I'd post links, but typing "Buffalo NAS throughput" in google comes up with multiple hits of reviews complaining about throughput.

    --
    I tried every decent and legal way I could think of to resolve the issue w/the business before I rented the chicken suit
    1. Re:Not Buffalo by frank_adrian314159 · · Score: 1

      I second this. The Buffalo units have a reasonably good UI and are easy to manage, but they are hideously slow.

      --
      That is all.
    2. Re:Not Buffalo by dave562 · · Score: 2, Informative

      I agree with the suggestion to avoid Buffalo. Someone else responded to this thread and said that their UI is good. My experience was just the opposite. The UI sucked and trying to get the thing integrated into Active Directory was a nightmare. The setup appears to be straight forward. Specify domain name, specify domain username/password combo. The reality of the situation turned out to be decidedly different and required numerous calls to tech support, firmware updates and a lot of headaches.

    3. Re:Not Buffalo by Anonymous Coward · · Score: 0

      Oh, it's not even a matter of CPU power! These slow NASes will save a miniscule amount on the network chip..

                With a compatible chip, samba or whatever calls sendfile(), and the linux kernel coordinates DMA transfers straight from the hard disk to the ethernet transmit buffer. Since this is all going to be interrupt-driven the CPU is essentially idle when a file is transferring.

                Without that, it's a IDE disk read into some buffer in memory, calculate checksum, copy it to the network card. The checksum and copy to network card are both entirely done by the CPU. You'd have about 125 million bytes a second going through at full gigabit speed, and 400 million instructions a second to deal with it.. (400mhz, MIPS and the like aren't usually pipelined). It's just nowhere near enough.

                This was one secret of the old IBM mainframes -- the CPU was a real clunker until recently, but the design made sure that the CPU was *only* computing, not dicking around with peripherals.

    4. Re:Not Buffalo by repvik · · Score: 1

      Try the Linkstation Pro, it's got a 400mhz cpu. With custom firmware (eg. from http://foonas.org/) it does actually deliver.

      With stock firmware, ANY soho NAS sucks. Really. They take a dev-board from a CPU-company, modify it a wee bit, then take the sdk and slap on a webinterface and samba. Some haphazardly applied superglue later, a finished product.
      I've been hacking a lot of NAS-devices, and I've yet to see one that has *any* performance optimizations done.
      One example is the QNAP TS-101 (now obsolete), where disk read speeds were ~15MB/sec in the original firmware. With a new kernel and rootfs, read speeds were boosted to ~32MB/sec.

    5. Re:Not Buffalo by DesertBlade · · Score: 1

      I don't think that is an issue with Buffalo but AD. I have had numerous issues integrating things into AD especially windows AD.

      --
      Half of writing history is hiding the truth.
    6. Re:Not Buffalo by rtobyr · · Score: 1

      Other than the fact that it's running a wimpy ARM processor, my biggest beef with the Buffalos is this: I bought three Terastation II Rackmount units (2gb each). Soon after moving my files to them, they started to crash. Upon calling Buffalo tech support, they informed me (but refused to provide documentation), that these NAS units perform well with a fair number of large or medium files; however I was hosting hundreds of thousands of small files, which causes problems. Supposedly, I'm only one of a "handful of customers" that has this problem.

  13. Mac Mini by Anonymous Coward · · Score: 0

    If you're going to buy a NAS we've found the QNAP ones give around 30 to 40 mbps

    however

    if you have os x clients if you use an os x machine then they will be able to search the drive with spotlight. any other operating system and they will have to do a normal slow search.

    1. Re:Mac Mini by alexandre_ganso · · Score: 1

      I second that. It works, it is way faster than those nasty NAS with crappy linux distributions fitted into a 400mhz processor, and you can simply attach external usb drives for more space in hot-swap.

      Besides, the external HDs may fit as a time machine.

      Have you though about the apple's Time Capsule?

    2. Re:Mac Mini by Anonymous Coward · · Score: 0
  14. Addonics NAS - usb connector by Anonymous Coward · · Score: 0

    Addonics NAS is a $55US device with a USB on one side and RJ45 on the other. I dont know anything about it other than it was written up recently on Gizmodo

    "For only $55, Addonics claims that this tiny gadget can easily turn any USB storage device into a full-fledged Network Attached Storage (NAS) server with support for both SMB and FTP access."

    1. Re:Addonics NAS - usb connector by Gary0G · · Score: 0

      From the Addonic FAQ: "The NAS adapter supports only FAT32 file system at this time." Back to shopping for a NAS...

    2. Re:Addonics NAS - usb connector by larry+bagina · · Score: 1

      You can buy an NSLU2 for $5 less and get 2 USB ports.

      --
      Do you even lift?

      These aren't the 'roids you're looking for.

    3. Re:Addonics NAS - usb connector by repvik · · Score: 1

      And, the largest NAS-hacking community providing alternative firmware and package feeds: http://www.nslu2-linux.org/

  15. Infrant (now netgear is awesome) by Wild_dog! · · Score: 1

    Infrant Ready NAS+ is very nice. I have had it on my gigaE network with OSX Ubuntu and Windows XP MCE as well as various other machines and laptops running win2k and XP professional. It is very fast and snappy with gigaE. I had been running it on a slower network and it was still able to reliably stream video content and simultaneously do large file transfers without a hiccup. With the HP procurve gigabit ethernet switch it is all just much faster. I am not a technical user, but it has been very reliable with no problems and seems very very speedy. Big thumbs up for this option.

  16. Synology by Anonymous Coward · · Score: 0

    Synology has a fair amount of good data on their website, and I've found the interface to be nice, though I'm a lite user. They're one of the relatively few companies that I've read about that has consistently good reviews. Their items are pricey, however.

    1. Re:Synology by repvik · · Score: 1

      The Synology devices are notoriously hard to hack if you decide you want an alternative firmware at a later point. But I guess that is what is to be expected from Microsoft culture (Synology was started by ex-MS employees).
      I'd go with QNAP instead. Pretty much the same products (and price range...), but with a community backing.

  17. OpenSolaris / ZFS by msdschris · · Score: 3, Informative

    Build it yourself and install Opensolaris. ZFS rocks.

    1. Re:OpenSolaris / ZFS by Anonymous Coward · · Score: 0

      I agree. Or look at using Nexenta

    2. Re:OpenSolaris / ZFS by myz24 · · Score: 1

      I was looking for an OpenSolaris post. I would also recommend OpenSolaris. Install it and with one command you can share out a bit of space via smb.

    3. Re:OpenSolaris / ZFS by Anonymous Coward · · Score: 0

      ZFS may but Solaris pretty much bites the big one.
      The most unfriendly OS I've ever had the misfortune to use.
      You think that's a good selling point? No wonder they are giving it away!!

    4. Re:OpenSolaris / ZFS by elvisior · · Score: 1

      How come this only gets a 3?

      I use netapp's at work and for anything where I didn't have an unlimited budget I'd use opensolaris with zfs.

      For a turn key solution based on the same technology perhaps you should check out nexenta which is a commercialisation of opensolaris and zfs to solve the problem at hand here...

      No I don't work for them.. but I've used it and it's cool. :)

      http://www.nexenta.com/corp/

    5. Re:OpenSolaris / ZFS by Anonymous Coward · · Score: 0

      This is what i've done and w/ great success. Open solaris and afs allow you to not worry about some of the data protection issues that you'd normally have to worry about w/ multidrive nas.

      http://searchsmbstorage.techtarget.com/generic/0,295582,sid188_gci1327343,00.html

  18. Timely by PingXao · · Score: 1

    I'm in the same boat as of last week. I have 2 machines, an XP box and a Linux server, both with GbE adapters from different manufacturers (Intel and Broadcom). Last week I decided to reduce a few TV programs I had captured using the firewire connection from my cable box. One was about 12 GB. Nothing like transferring a 12 GB file to clue you in to the fact that the little router you THOUGHT was doing 100 Mbps is actually only doing 10 Mbps. What a nightmare. So now I'm also looking for something that can support my GbE adapters. The router I can replace, but even if it did handle 100 Mbps traffic, even that would be too slow for these files.

    1. Re:Timely by Fulcrum+of+Evil · · Score: 1

      So go get a GigE switch and use the router for cable only - they aren't that expensive these days.

      --
      "We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
  19. Buffalo Tech Mac compatibility by Gizzmonic · · Score: 2, Informative

    I have a Terastation 2 (by Buffalo) and I am plugged into 100Mbps ethernet at work, so I can't tell you about the throughput, but I can tell you that the Terastation Mac stuff is very half-assed. I couldn't get AFP/Appletalk to work at all and while SMB is rock solid for large files, it cannot handle large amounts of small files. It chokes on directories with huge amounts of files (not sure if that's a limitation of the Finder or the Terastation's fault, though). I had a user's backup program run amok and generate millions of tiny .tmp files over the course of about a month, and I was unable to delete them from OS X, even when waiting days. I had to use Windows Explorer, which was slow but eventually worked.

    The built-in webpage used for administration is pretty terrible too. It works best with IE 6 on Windows, but even with that, sometimes the columns don't line up properly. If you misclick, you could end up changing the wrong shared folder.

    On the plus side, the Terastation 2 is pretty cheap. I'd give it about a B minus in terms of what I need it to do.

    --
    (-1, Raw and Uncut is the only way to read)
    1. Re:Buffalo Tech Mac compatibility by geekmansworld · · Score: 1

      I second regarding Buffalo's half-assedess.

      Our organization has been using a Terastation for a few years now. While its generally a solid product for basic usage, it becomes difficult to work with when attempting any particularly complex configuration. And don't ask the Buffalo support staff for help, they don't know anything about the backend of their product.

      If you're looking for flexibility, I'd recommend ditching the NAS idea entirely and going for a basic file server.

    2. Re:Buffalo Tech Mac compatibility by richard_weller · · Score: 1

      I recently got a LinkStation Pro Duo 1TB for backups at work - its very slow. A Duron based box (1Ghz) with 2, 3-4 year old IDE disk's transfers data nearly twice as fast over gigabit.

      I can't get it to store timestamp's properly either.

      Although cheap, I would avoid Buffalo NAS boxes.

  20. I got a linksys NSLU2 by Nursie · · Score: 1

    Great little debian server, really bad performance as a NAS. Even with Debian on there.

    I like the idea of the QNAP Turbo stations - effectively a modernised NSLU2 with 256 MB of RAM and a 500MHz chip, but then I want another server rather than an actual NAS...

    1. Re:I got a linksys NSLU2 by larry+bagina · · Score: 1

      I've been looking at the QNAP 409Pro for a while (I wish the price would drop on it a bit). The limited memory on my slug prevents me from doing a lot of fun stuff with it :/

      --
      Do you even lift?

      These aren't the 'roids you're looking for.

    2. Re:I got a linksys NSLU2 by Nursie · · Score: 1

      I fancied the 109 as it doesn't have a fan and it's small.

      The slug just about runs torrentflux-b4rt and ushare successfully, but I wouldn't want to try anything more. That's why I have the second one for web/mail/ssh.

    3. Re:I got a linksys NSLU2 by larry+bagina · · Score: 1

      correction: just the regular QNAP 409 (I thought the pro had more memory). The $100 difference is NFS support, which doesn't matter since the first thing I'll do is put debian on it.

      --
      Do you even lift?

      These aren't the 'roids you're looking for.

    4. Re:I got a linksys NSLU2 by dropadrop · · Score: 1

      I have a qnap TS-109 and it has worked well. You get root shell without hacking the device and can setup all the stuff you lose on the cheaper version yourself (nfs etc). I think the biggest problem with it is slow replication for backups. I had to write my own scripts to back it up on an external device.

      The best thing about this device is how silent it is. There are no fans, and using a silend disk (I have a 1TB wd green power which is pretty silent) you can only hear the sound of the disk rotating if you put your ear next to it.

    5. Re:I got a linksys NSLU2 by repvik · · Score: 1

      The TS-109 will probably run faster if you run foonas on it ;)

  21. NAS Charts by Anonymous Coward · · Score: 2, Informative

    www.smallnetbuilder.com maintaines a NAS Chart, I find it quite complete and recent.(http://www.smallnetbuilder.com/component/option,com_nas/Itemid,190/)

    1. Re:NAS Charts by fool · · Score: 1

      this site (smallnetbuilder) is the best i found doing research on similar topics recently. i am less concerned with bandwidth so was more interested in what's behind the curtain and featuresets which is covered well, but they certainly also have some bandwidth charts, including performance w/use of jumbo frames (which sounds like it might be of interest to you).

    2. Re:NAS Charts by Anonymous Coward · · Score: 1, Insightful

      This implies to me "make your own NAS" in answer to the original question. The top ones were 50MB/sec in general (the one showed 75MB writes but wasn't even up to 50 reads.. odd results indeed.) You can get pretty high read/write speeds to any recent disk. On a traditional PCI system, make sure the disk controller and gigabit are not on the same bus, gigabit's enough to max a PCI bus. I would balance these out for a PCI Express system as well but it shouldn't be as crucial. If you get a fast enough disk you should be able to max out a gigabit link, if the remote machine is fast enough to handle it.

                Software? For ease of administration, I would either 1) Find some distro that just runs a NAS. 2) Put on Ubuntu, you go to some admin menu and turn on file sharing (samba) and file sharing (nfs). Not literally "out of box", but easy enough so you could screenshot every step and have literally anyone do it. This helps persuade people who want out-of-the-box but can't have it, that it won't take some awful voodoo just to recreate the box later.

  22. personal Drobo experience by jurv!s · · Score: 1

    Drobo v2 connected via FW800 maxes out at 50MBps reads and ~35MBps writes.

    --
    sigs are for fools and trolls. no signature is *always* appropriate. you should turn them off in your preferences.
  23. Go to SmallNetBuilder.com by sco_robinso · · Score: 3, Informative

    They have the most comprehensive benchmarks and NAS's around (that I've stumbled across, at least). Also, lots of good tests showing various things like Jumbo frames, etc. Very good overall.

    I frequent the site a bit, and there's a couple tricks to getting good performance out of a NAS, or LAN throughput in general.

    1. Use Jumbo Frames, period.
    2. Use PCI-e NIC's, onboard or PCI just can't deliver the speeds offered by GigE. You can find smiple intel PCI-e nics for under $20.
    3. Drives make a big difference, obviously.

    www.smallnetbuilder.com -- Good site.

    1. Re:Go to SmallNetBuilder.com by setagllib · · Score: 1

      I've benchmarked onboard and PCI NICs and get over 850Mb/s throughputs with iperf and netperf. Sure you could get another 50-100Mb/s with PCI-e, but that's practically a rounding error.

      --
      Sam ty sig.
    2. Re:Go to SmallNetBuilder.com by Propaganda13 · · Score: 1

      I notice someone else pointed out Tom's Hardware for a review. Tim Higgins was part of smallnetbuilder and Tom's Networking. I know he is still reviewing for smallnetbuilder, but I'm not sure about the Tom's part.

    3. Re:Go to SmallNetBuilder.com by Anonymous Coward · · Score: 1, Interesting

      My onboard Realtek PCI-E GigE ethernet chip can consistently deliver 999mbit in benches (using iperf or others).

    4. Re:Go to SmallNetBuilder.com by Anonymous Coward · · Score: 0

      That's not surprising since being onboard does not alone indicate if the NIC hangs off the PCI or PCI-E bus. Many chipsets can spare an x1 for that purpose.

    5. Re:Go to SmallNetBuilder.com by troll8901 · · Score: 1

      1. Use Jumbo Frames, period.

      Are there any compatibility issues with Jumbo Frames?

    6. Re:Go to SmallNetBuilder.com by OneMadMuppet · · Score: 1

      >1. Use Jumbo Frames, period.

      To use Jumbo Frames every NIC and switch must be configured to have the same MTU. If you try to use Jumbo Frames on a network with 100Mb/s cards (or wireless) things will break in interesting ways, and your network will actually run much slower.

      >2. Use PCI-e NIC's, onboard or PCI just can't deliver the speeds offered by GigE.

      These days most onboard NICs are PCI-E / GigE.

    7. Re:Go to SmallNetBuilder.com by Anonymous Coward · · Score: 0

      There are enough mainboards with onboard NICs that are connected via PCI-E. Maybe not the 'consumer' mainboards, but among the 'server' mainboards, even the entry level ones, this is rather common.

    8. Re:Go to SmallNetBuilder.com by setagllib · · Score: 1

      More importantly, it's just not that big a deal. GigE performance on a good card is still limited by software, such as buffer sizes, MTU config, network stack implementation, application implementation, etc. but of course 10GigE is definitely a PCI-e only endeavour.

      --
      Sam ty sig.
    9. Re:Go to SmallNetBuilder.com by drsmithy · · Score: 1

      2. Use PCI-e NIC's, onboard or PCI just can't deliver the speeds offered by GigE. You can find smiple intel PCI-e nics for under $20.

      It's not at all unusual these days for onboard NICs to be hang directly off the chipset, which is going to give you far better performance than a PCIe-based card.

  24. Anonymous by Anonymous Coward · · Score: 0

    Check Tom's Hardware. They reviewed the Synology DS207+ this week and have benchmarks for similar products for you to compare.

  25. How automated is your testing? by syousef · · Score: 2, Insightful

    If your testing is highly automated, I can't help you as I don't have a lot of experience with high speed networking.

    If your testing is reasonably manual, consider storing your data set on removable hard drives which are manually plugged into one computer, data is copied, then disconnected and moved to the other. A USB 2 interface will give you the most compatibility given the wide variety of hardware you're using, but perhaps there may even be hardware that does hot plugging E-SATA properly if you're willing to pay a premium.

    Remember, for really high bandwidth physical media being shipped from one location to another is still a solution which should be considered.

    --
    These posts express my own personal views, not those of my employer
    1. Re:How automated is your testing? by ceoyoyo · · Score: 1

      Decent Gig-E to a decently configured RAID box will wipe the floor with USB 2.

      Shipping physical media is often a good alternative to a WAN or the Internet, but isn't nearly as practical when you put it up against a gigabit LAN.

  26. Refund policy? by arootbeer · · Score: 1

    The nice thing about hardware like these is you can call up newegg, order one, play with it for 15 days, and wipe and return it if it doesn't fit your needs. I don't know exactly what your timeline is (from the question it doesn't sounds particularly time-sensitive), but unless you need the last word today, just buy the different ones you're thinking about and try them out.

  27. Build it yourself by kosmosik · · Score: 1

    Off-shelf NAS device will be not only slow but also full of various bogus bugs with which you need to wait for vendor to issue firmware update...

    Just build it yourself - build a PC. You have plenty of options:

    1. If you have a rack somewher buy a low end rack 2U rack server with enclosures for SATA disks and some decent RAID controller.

    Or:

    2. Build yourself a PC in tower enclosure. Get some Core 2 Duo mobo (cheapest), medicore ammount of RAM - SMB and NFS and AppleTalk servers with Linux operating system will eat up something like 80MB for the system and 10MB per client computer - go figure, the rest of RAM is for I/O buffers. Stuff as much as you can get SATA disks into that (like 4x 1TB). Setup it with software RAID. And you are done with it. Probably it will be much cheaper than decent NAS box (so called SoHo boxes are no worth even looking at).

    Do so and you have a decent storage that is more efficent that your network.

    You said about network efficency? Well - this has nothing to do with NAS box. You can have the best performing NAS box - but if your network is weak - well here goes your efficency.

    So as for network buy managable switch that can cope with Linux channel bonding - with that you can bond N ethernet channels and get network transfers somewhat lower than N*interface speed.

  28. DIY is Your Best Bet by mpapet · · Score: 1

    Most of the embedded-style NAS will crap out WAY below whatever throughput you are looking for.

    The trick is going to be maxing out the transfer bandwidth by identifying the bottlenecks in a Linux file server.

    The most direct route I can imagine is a proper SAN and fibre channel controllers. Not cheap, but my time isn't either.

    --
    http://www.maxineudall.com/2010/02/should-economists-be-sued-for-malpractice.html
    1. Re:DIY is Your Best Bet by MyNameIsMike · · Score: 2, Insightful

      Just my $0.02: I have been running my server (named "JUPITER") SMB + Apache + Webmin on Ubuntu 6.06 LTS with LVM and RAID on an old Compaq Dual Processor SP750 with 256MB of RAM and a few 500 GIG PATA disks for over 2 years now. (Ran the same hardware under Fedora before) . The Network is an old 100/10.

      Stability is SUPERBE --- system has NEVER crashed --- only downtime is when the power goes out or I go on vacation.

      Speed is satisfactory --- everyone seems happy with the network. It just works.

      Compatibility is GREAT --- 1 Windows VISTA , 1 Windows XP, 3 MacOSX and 1 Ubuntu 8.04 machines all use it. (even my DD-WRT based router has a share on the server.)

      Cost is ridiculously low --- I probably couldn't give the hardware away without paying someone to take it... it's that old.

      I've been wanting to upgrade my server to something newer and sexier, (why? ... because it's newer and sexier...) but my old system refuses to die, and it's so stable that I don't want to go through the trouble of an upgrade. FYI

  29. Thecus N5200 Pro NAS.. by hunter44102 · · Score: 1

    http://www.thecus.com/ has a whole line of NAS's from SOHO to Enterprise. I believe they are linux based.

    1. Re:Thecus N5200 Pro NAS.. by am+2k · · Score: 1

      Yes, they're Linux-based. I've just installed an ssh server on my N3200, now I have a very cheap hackable Linux box with space for 3 disks.

      There's a Wiki for more information about hacking Thecus products.

      Btw, they're using netatalk for Mac-Support, which appears to work really well.

    2. Re:Thecus N5200 Pro NAS.. by hunter44102 · · Score: 1

      I have the N3200 also. Will have to check out the Thecus hacks. Thanks.

    3. Re:Thecus N5200 Pro NAS.. by am+2k · · Score: 1

      Note that you have to unzip the files there ONCE to get the .mod-file to upload to the device (which is really a .tar.gz-file). Took me a while to figure that out.

  30. Drobo makes a great product by analogueblue · · Score: 1

    I've been VERY pleased with the reliability and speed of my Drobo. If you don't want to spend the time rolling your own, dealing with linux raid drivers and related issues, etc... Just get a Drobo.

    1. Re:Drobo makes a great product by Mendy · · Score: 1

      It's certainly a very "shiny" product but the speed is average even amongst USB external enclosures. The supplied software is also rather quirky; the way it works is that it claims to always have a 2Tb partition and then fudges the amount of free space accordingly. The monitoring software also seems to require being run as an administrator.

      One other thing about them which amused me was that whilst they support hot removal of drives the lack of caddies has meant they've had to put warning stickers advising you to turn the unit off for 15 minutes before doing so to avoid being burnt. Nice.

  31. Roll your own by Phred+T.+Magnificent · · Score: 2, Insightful

    Your best performance is likely to come by rolling your own. Off the shelf SOHO devices are built for convenience, not throughput.

    Grab a PC (need not be anything top-of-the-line), a good server NIC, a decent hardware RAID card (you can usually get a good price on a Dell PERC SATA RAID on ebay), and a few SATA hard drives. Install something like FreeNAS or NexentaStor (or, if you want to go all the way, FreeBSD or Linux and Samba).

    --
    Where is the wisdom we have lost in knowledge?
    Where is the knowledge we have lost in information?
  32. SmallNetworkBuilder or DIY by Anonymous Coward · · Score: 0

    Just visit small network builder and look up their NAS charts. Contains benchmark for every device they ever tested.

    Building your own is still strongly recommended. If you want a quality system, buy a dirt cheap HP Proliant ML115 server and four fast >1TB SATA II drives. Use software raid 10 mode (mdadm can handle this) and it'll fly. It's also reusable, I just converted mine to run whitebox VMWare ESXi without any hassles.

  33. UnRaid: when build-from-scratch isn't fast enough by Overzeetop · · Score: 3, Informative

    Okay, unRaid is not particularly fast compared to an optimized system, but it's expandable, had redundancy, is expandable, is web managed, plays nice with windows, sets up in about 20 minutes, costs $0 for a three disc license and $69(?) for a 6 disk license.

    My total unoptimized box on an utterly unoptimized Gb network (stock cards, settings, with 100 and 1000 nodes) and unmanaged switches just transferred an 8.3GB file in a hair under three minutes. From a single, cheap SATA drive to a Vista box with an old EIDE drive. Now 380Mb/s is not blazingly fast, but remember that it took almost no effort.

    http://lime-technology.com/

    No connection except as a happy customer with a 4TB media server that took longer to assemble the case than to get the SW running. If only my Vista Media Center install has been this easy.

    --
    Is it just my observation, or are there way too many stupid people in the world?
  34. ReadyNas NV+ by davew666 · · Score: 1

    I have one of these with 4x750GB seagate drives. I get 35mb/sec read and about 15-20 write. I'm using OSX on AFS and SMB shares

  35. Look at the Promise SmartStores by Anonymous Coward · · Score: 0

    They support all platforms, including the Apple networks, include RAID. The four drive one even has RAID 5 support.

    You'll have to buy drives for the unit, but they are very nice and perform well. Try and stay away from the media sharing add-ons for it, they'll lock the unit up from time to time.

  36. You don't want software raid and the (cheapest) MB by Joe+The+Dragon · · Score: 1

    You don't want software raid and the (cheapest) MB sucks way out of data chipset also you don't have to have on board vidoe as it takes up system ram and chip set i/o even if you are useing the system for much.

    you will want 2gb or more ram + dual gig-e or more in teaming + some kind of a raid card a good pci-e x4 or better one is about $250+

  37. Depends on how many simultaneous dataset transfers by Anonymous Coward · · Score: 0

    The type of NAS will depend on how many simultaneous transfers. It wasn't said if the large datasets were being transferred sequentially or in parallel. ATA does not handle simultaneous IO as well as SCSI/FC. If they are sequentially, a NAS with ATA disks would be sufficient and RAID3 might be more optimal. If there are many simultaneous transfers then fiber channel drives and a more expensive NAS might be needed.

  38. NAS disk architecture by anegg · · Score: 5, Interesting

    If you use a single disk NAS solution and you are doing sequential reads through your files and file system, your throughput can't be greater than the read/write speed of a single disk, which is no where near GigE (1000 Gbps is about 125 MB/second ignoring network protocol overhead). So you will need RAID (multiple disks) in your NAS, and you will want to use striped RAID (RAID 0) for performance. This means that you will not have any redundancy, unless you go with the very expensive striped mirror or mirrored stripes (1+0/0+1). RAID 5 gives you redundancy, and isn't bad for read, but will not be that great for writes.

    As you compare/contrast NAS device performance, be sure that you understand the disk architecture in each case and see oranges to oranges comparisons (i.e, how does each one compare with the RAID architecture that you are interested in using - NAS devices that support RAID typically offer several RAID architectures). Also be sure that the numbers that you see are based on the kind of disk activity you will be using. It doesn't do much good to get a solution that is great at random small file reads (due to heavy use of cache and read-ahead) but ends up running out of steam when faced with steady sequential reads through the entire file system where cache is drained and read-ahead can't stay ahead.

    Once you get past the NAS device's disk architecture, you should consider the file sharing protocol. Supposedly (I have no authoritative testing results) CIFS/SMB (Windows file sharing) has a 10% to 15% performance penalty compared to NFS (Unix file sharing). I have no idea how Apple's native file sharing protocol (AFP) compares, but (I think) OS X can do all three, so you have some freedom to select the best one for the devices that you are using. Of course, since there are multiple implementations of each file sharing protocol and the underlying TCP stacks, there are no hard and fast conclusions that you can draw about which specific implementation is better without testing. One vendor's NFS may suck, and hence another vendors good CIFS/SMB may beat its pants off, even if the NFS protocol is theoretically faster than the CIFS/SMB protocol.

    Whichever file sharing protocol you choose, its very possible it will default to operation over TCP rather than UDP. If so, you should pay attention to how you tune your file sharing protocol READ/WRITE transaction sizes (if you can), and how you tune your TCP stack (windows sizes) to get the best performance possible. If you use an implementation over UDP, you still have to pay attention to how you set your READ/WRITE buffer sizes and how your system deals with IP fragmentation if the UDP PDU size exceeds what fits in a single IP packet due to the READ/WRITE sizes you set.

    Finally, make sure that your network infrastructure is capable of supporting the data transfer rates you envision. Not all gigabit switches have full wire-speed non-blocking performance on all ports simultaneously, and the ones that do are very expensive. You don't necessarily need full non-blocking backplanes based on your scenario, but make sure that whatever switch you do use has enough backplane capacity to handle your file transfers and any other simultaneous activity you will have going through the same switch.

    1. Re:NAS disk architecture by swillden · · Score: 1

      your throughput can't be greater than the read/write speed of a single disk, which is no where near GigE (1000 Gbps is about 125 MB/second ignoring network protocol overhead)

      Most bog-standard SATA drives should be able to push over 100 MBps on a sequential read. I'd go RAID0 to help with the case when files are a bit fragmented and other activity is going on, but under ideal conditions a single drive should be able to very nearly saturate a Gig-E TCP stream.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    2. Re:NAS disk architecture by Anonymous Coward · · Score: 0

      Good reviews found here:

      http://www.smallnetbuilder.com/component/option,com_nas/Itemid,190/chart,13/

    3. Re:NAS disk architecture by Anonymous Coward · · Score: 0

      Also, make sure your network can make use of jumbo frames.

    4. Re:NAS disk architecture by jon3k · · Score: 1

      I can't believe in this day and age people would still recommend raid0.

      http://www.anandtech.com/storage/showdoc.aspx?i=2101

  39. Infrant's ReadyNAS forums by Anonymous Coward · · Score: 0

    If you check the forums on infrant's website you can get lots of real-world through-put numbers from users. I have the ReadyNAS and am quite happy with it.

  40. Go virtual by Anonymous Coward · · Score: 0

    If you have the money, why not go virtual? Put all the different operating systems on one box using VirtualBox, then they can all share the same hard drive.

    This results in no reliance on network throughput, just disk read/write time.

    Instead of migrating between boxes to get different environments, migrate between operating systems on the same box.

  41. Network won't be your bottleneck. by m0e · · Score: 5, Informative

    Disk will always be. Since disk is your slowest spot you will always be disk I/O bound. So in effect there's no real reason to worry about network throughput from the NIC. NICs are efficient enough these days to just about never get bogged down. What you would want to look at for the network side would be your physical topology -- make sure you have a nice switch with nice backplane throughput.

    About disks:

    Your average fibre channel drive will top out at 300 IO/s because few people sell drives that can write any faster to the spindle (cost prohibitive for several reasons). Cache helps this out greatly. SATA is slightly slower at between 240-270 IO/s depending on manufacturer and type.

    Your throughput will depend totally upon what type of IO is hitting your NAS and how you have it all configured (RAID type, cache size, etc). If you have a lot of random IO, your total throughput will be low once you've saturated your cache. Reads will always be worse than writes even though prefetching helps.

    If you're working with multi-gigabyte datasets, you'll want to increase the number of spindles (ie number of disks) to as high as you can go within your budget and make sure you have gobs of cache. If you decide to RAID it, which type you use will depend on how much integrity you need (we use a lot of RAID 10 with lots of spindles for many of our databases). That will speed you up significantly more than worrying about the NICs throughput. don't worry about that until you start topping a significant portion of your bandwidth -- for example, say 60MB/sec sustained over the wire.

    This doesn't get fun until you start having to architect petabytes worth of disk. ;)

    1. Re:Network won't be your bottleneck. by bunny.rabbit.3 · · Score: 1

      Wow. Just. Wow. Somebody knows their hard disk architecture.

    2. Re:Network won't be your bottleneck. by m0e · · Score: 1

      Welcome to real-time processing IT. Where you either know your shit soup-to-nuts or go home crying to mommy. :D

    3. Re:Network won't be your bottleneck. by Anonymous Coward · · Score: 2, Informative

      It sounds like you do this as your day job working with big expensive NAS and SAN equipment. Yes, in those environments you'll be I/O bound long before you're disk-bound or NIC bound. Sadly, the SOHO equipment is far, far worse. By and large, their throughput ranges from sad to atrocious. See SmallNetBuilder's NAS Charts for some benchmarks that will make you weep.

    4. Re:Network won't be your bottleneck. by afidel · · Score: 1

      Actually the network CAN be the bottleneck even at low throughput. We had an Oracle server that saw an ~20% decrease in CPU load and significant reduction in latency by going with a TOE card, a $50K server (running several times that in software) sped up significantly with the addition of an $800 card, the best bang for the buck I have seen in over a decade in IT.

      --
      There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
    5. Re:Network won't be your bottleneck. by Fulcrum+of+Evil · · Score: 1

      We're talking file servers here - a DB has some other obvious heavy loads.

      --
      "We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
    6. Re:Network won't be your bottleneck. by drsmithy · · Score: 1

      Actually the network CAN be the bottleneck even at low throughput. We had an Oracle server that saw an ~20% decrease in CPU load and significant reduction in latency by going with a TOE card, a $50K server (running several times that in software) sped up significantly with the addition of an $800 card, the best bang for the buck I have seen in over a decade in IT.

      How did you manage to get a $50k server that didn't already have a TOE card in it in the first place !? Heck, even bottom of the barrel 1U servers have had TOE NICs in them for years.

    7. Re:Network won't be your bottleneck. by afidel · · Score: 1

      The server was built summer of 2006, there were no servers with a dedicated TOE card (not checksum offloading but full stack offloading). Granted today all of HP's servers ship with them, but you must be running Windows with either the Advanced Networking Pack (doesn't run so well on x86, much better on x64) or SP2.

      --
      There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
    8. Re:Network won't be your bottleneck. by Anonymous Coward · · Score: 0

      " don't worry about that until you start topping a significant portion of your bandwidth -- for example, say 60MB/sec sustained over the wire."

      60MB/s loads over the wire are trivial. In fact when you're talking about bulk transfer, 60MB/s is absolutely squat. Even a cheap, bargain basement NAS built out of a scratch server and a mere two channel, 16 disk PowerVault DAS will be NIC-bound. Your underlying disk subsystem will be capable, at a minimum, of 300MB/s mixed read-write, and you're talking about being surprised by a 60 meg load?

      Hell, sitting at home holding music, movies, and backup images of the wife's laptop I have a 4-spindle SATA array that can sustain over 100MB/s writes, and even that is bound by wire speed for bulk operations.

      I can saturate a port of gigE with a single VMotion operation. I can knock out two ports worth of bandwidth just by deploying a VM from a template. A sane network storage solution is going to have 4 or more ports into the core switch merely because it actually IS incredibly easy to saturate.

      Just because you run IOpS-bound transactional loads doesn't mean the world does. Just because you consider 60-70MB/s a high water mark doesn't mean the world does.

      Those of us who do large bulk transfers absolutely, positively MUST worry about NIC throughput because gigE is REALLY REALLY SLOW.

    9. Re:Network won't be your bottleneck. by rthille · · Score: 1

      Network won't be your bottleneck

      You address this in your post, but your subject's assertion of course depends on your workload and how many spindles and cache you throw at the problem. I'd guess a Sun box with 48 drives and 64GB of memory could keep your network plenty busy, especially if you are just serving up say large archives of video editing projects to local storage @ workstations where all the seeking is going to happen. If you've got less disk than memory, then a "devil's advocate" program could always limit you to the thruput you can get from any individual disk having to seek to any particular block. But in the real world programs aren't like that, and you'll benefit from cache and multiple clients.

      (email servers, which I work on, are almost as bad as databases wrt disk loading... :-)

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    10. Re:Network won't be your bottleneck. by Anonymous Coward · · Score: 0

      riiiiiiiiight. tell me all about this multiple petabyte storage system you architected.

  42. Buffalo Linkstation by Anonymous Coward · · Score: 0

    Hi,
          I get 15 megabytes a second out of my Bufallo Linkstation live 500 Gb. The bottlneck is (I believe) the relitavely slow ARM chip in the Linkstation. For comparison, I get 6 megabytes/s when talking to the linkstation from a fast Ethernet (as opposed to gigabit) device.

    That measurement is transferring large files. It is a bit slower for many small files.

    Jim.

  43. Disk Speed Is Your BottleNeck by Anonymous Coward · · Score: 0

    The problem with off the shelf NAS products is two fold: the linux installs aren't disk optimized and the disks are usually the slowest (eg - cheapest, eg - highest profit margin) the company can find.

    Your best bet, hands down, is to build your own. But, at best, your "bursty" speed is going to max out around 300 - 4000 Mbps and long term transfer will approach, at best, around 40 Mbps if you're lucky. If you setup a mirrored and striped raid array, you might be able to come close to doubling that. If you get some really expensive 10k RPM disks and a high end RAID controller, you can probably double that.

    Keep in mind that's for single transfers. Once you add in multiple people trying to do IO on a single controller, your performance will take a hit.

    With that said, I have a first gen Buffalo LinkStation. I love it, it does exactly what I need. But, as everyone else has said, it is DOG SLOW. My brother used it without problem from OS X 10.3 - 10.5.

    1. Re:Disk Speed Is Your BottleNeck by afidel · · Score: 1

      Modern HDD's can easily do 50MBps sustained for bulk transfer so even a 3 drive RAID5 with a CPU and network card that can keep up should be able to saturate a Gig-e network without a problem.

      --
      There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
  44. We run the ReadyNAS here at the office... by Anonymous Coward · · Score: 0

    We have ours built out to 2GB with RAID mirroring.

    Frankly the throughput and directory browsing speeds are disappointing. We are using 7200RPM 32MB cache drives. Large file copies or writes of large amounts of small files seem to take abnormally long, and our network is not congested in general.

  45. Build it yourself (with a kit!) by Anonymous Coward · · Score: 0

    Just an example - http://www.mini-itx.com/2008/12/08/via-launch-artigo-a2000-barebones-storage-server

    There are lots of 'kits' you can buy so you don't really have to do much.

    OpenFiler and FreeNAS work really well. Most SoHo systems don't have great throughput, and are substantially more expensive. Even the DroboApps can't compete with just installing something yourself if you want to download torrents (like satellite imagery) to it.

  46. Drobo != fast by Anonymous Coward · · Score: 0

    The Drobo is not fast.

  47. My old Buffalo Linkstation sucked so bad by Technomancer · · Score: 1

    I have built little PC do do my file serving.
    I dont know why they added gigabit port when the dinky CPU was barely capable of couple MB/s of IO.
    It was 250GB ATA drive and 266MHz PPC with 64MB of memory.
    My advice: build cheap small PC with Linux and software RAID. (I like software RAID better because you are not tied to particular RAID controller hardware so you can move it to newer system).
    It is going to be way faster than any embedded crap.

  48. Never underestimate the bandwidth.... by M0b1u5 · · Score: 2, Insightful

    Never underestimate the bandwidth of a guy carrying a bundle of removable hard drives around the office.

    Or a station wagon loaded with hard drives.

    Nothing can beat them.

    --
    How many escape pods are there? "NONE,SIR!" You counted them? "TWICE, SIR!"
    1. Re:Never underestimate the bandwidth.... by bunny.rabbit.3 · · Score: 1

      Ahh! The Google Sneakernet approach.

    2. Re:Never underestimate the bandwidth.... by Sponge+Bath · · Score: 1

      Nothing can beat them.

      Maybe a C-5 Galaxy full of hookers carrying hard drives.
      But that might be cost prohibitive for small office use.

    3. Re:Never underestimate the bandwidth.... by Anonymous Coward · · Score: 0

      Why you did not get at least 'funny' escapes me...

    4. Re:Never underestimate the bandwidth.... by Anonymous Coward · · Score: 0

      Or a semi-truck filled with 1.44" floppies ...

    5. Re:Never underestimate the bandwidth.... by Anonymous Coward · · Score: 0

      Great throughput but terrible latency, particularly across oceans.

    6. Re:Never underestimate the bandwidth.... by Jazz-Masta · · Score: 1
    7. Re:Never underestimate the bandwidth.... by Anonymous Coward · · Score: 0

      Actually, you need to use a Prius loaded with hard drives for the prior claim to be true.

  49. Thecus N2100 by opk · · Score: 2

    I've got an Thecus N2100 and the performance as a NAS isn't great. The CPU isn't powerful enough to take advantage of the gigE interface. For what you want, I'd get something more powerful which probably means an x86 box. For anyone who just wants a home server that doesn't consume too much electricity so can be left on all the time, a small ARM based box is great. I'm running Debian on it and it's really useful.

    1. Re:Thecus N2100 by jdowland · · Score: 1

      I'll second that. The n2100 can barely manage 10Mbps in my experience, despite having dual gigE NICs.

    2. Re:Thecus N2100 by magnosis · · Score: 1

      We've tried this device for a bit at the office, it was near catastrophic. We've lost all data that was on it after an expected drive failure. The box didn't work as advertised; as soon as one drive failed, the entire RAID became non-accessible and everything was corrupted before it even started to rebuild. Stay away from Thecus. Their tech support was mostly crap (except for 1 guy) and they refused to refund 11 days after purchase.

    3. Re:Thecus N2100 by Anonymous Coward · · Score: 0

      if you expected the drive failure , why was it catastrophic?

      Usually its the unexpected drive failures that are the problem.

  50. Yes you want software RAID and lots of memory. by Technomancer · · Score: 2, Insightful

    Actually, you dont want any RAID card, because it limits your upgrade and recovery options. Any modern CPU is not going to have any problems doing memcopy and XORing required for RAID.
    You do want as much memory as you can afford, especially that memory is cheap now.
    My little home server has 8GB of memory, it can sink huge write transfers very quickly. It uses 3 laptop SATA HDDs in RAID5 so it can take it's sweet time to write the data to HDD later because it effectively has 8GB disk cache.

    1. Re:Yes you want software RAID and lots of memory. by DamnStupidElf · · Score: 1

      I doubt if you're really getting the benefit from your cache that you think you are; and if you are, it's dangerous. A SAN/NAS should never cache data in memory and tell the client that its data is written to disk until the data has actually been written to some non-volatile storage. Expensive SAN/NAS devices have battery backed caches for exactly that reason. What good is your 8 GB of RAM if all the files you backed up into volatile RAM for the last few minutes are lost when the power goes out?

    2. Re:Yes you want software RAID and lots of memory. by TCM · · Score: 1

      Why would it take much time to write anyway? I get over 200MB/s write speed to a 5-component RAIDframe RAID5 using NetBSD.

      Write performance of RAID5 only sucks if you mess up stripe sizes and write block sizes of the upper layer. You want writes that don't require additional reads to compute the new checksum.

      --
      Of course it runs NetBSD. BTC: 1NT7QvbetmANwaMzhpVL6
    3. Re:Yes you want software RAID and lots of memory. by swb · · Score: 1

      What's the risk of cache loss, really? I mean, assuming something reasonable like a UPS? And what's the real risk of data loss vs. the vastly improved write performance of write back caching?

      I can't remember the last time I chose "write through" as an option, even when there was no battery backing.

    4. Re:Yes you want software RAID and lots of memory. by rthille · · Score: 1

      Exactly. I tried setting up my new NetBSD server with 4 drives in RAID5. Sucked eggs thru a straw. That's because 4 drives means 3 data blocks and 1 parity block. The filesystem (on NetBSD at least) can only do 2^n block sizes, and since 2 and 3 are relatively prime, you end up with _every_ write being a read-modify-write. It was 10x faster thruput to go to a 3+1 spare setup.

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
    5. Re:Yes you want software RAID and lots of memory. by SHaFT7 · · Score: 1

      seeing as you don't advocate a real raid card, you better make sure you don't have an issue with power, as a momentary dip will cause you to lose your cache, and therefore probably lose data and or go degraded/failed. even if you have a UPS, you better have usb/serial connectivity to enable the UPS to shutdown the machine.

    6. Re:Yes you want software RAID and lots of memory. by DamnStupidElf · · Score: 1

      OS crashes and hardware failure do happen. So do random acts of system administration like accidentally pulling a power cord. Additionally, when your UPS battery is running out, does your server have enough time to write 8 GB of cache when it determines that it's time to shut down? UPS's fail, too.

      It depends on your workload though. A database or other transactional storage without a write-through cache is basically a disaster waiting to happen, even with archive/replay logs, because the replay logs are probably still sitting in the cache when the database finally commits the transaction. You can utterly break a database that way, with no recourse except to fall back on the last full backup and any replay logs that are still good, while losing your ACID compliance because transactions are not guaranteed to flush to disk until a few seconds/minutes after an application has committed them.

      If you just have a bunch of movies or other simple files being stored, the danger is a little lower. You could potentially corrupt a few random files, which is bad, but it's not the same risk as an unrecoverable database. But any sort of processing system that's using the disk to store state information should really have write-through caching for safety.

    7. Re:Yes you want software RAID and lots of memory. by swb · · Score: 1

      You're assuming, though, that the cache flush happens really lazily, instead of the I/O board (raid card, or whatever) doing at top priority as soon as the underlying disk hardware is capable of committing the writes.

      I also think the trend is to more cached writes, not fewer of them, and often in ways you can't control the latency of the cached writes (think SANs or other remote storage where not having cached writes would mean huge latency penalties and unusable storage).

    8. Re:Yes you want software RAID and lots of memory. by DamnStupidElf · · Score: 1

      You're assuming, though, that the cache flush happens really lazily, instead of the I/O board (raid card, or whatever) doing at top priority as soon as the underlying disk hardware is capable of committing the writes.

      I'm assuming that the cache is actually used, e.g. if the original poster needed 8 GB of cache, then the workload was a ton of small writes that had to be collected in the 8 GB cache so they could be written in huge chunks and max out the throughput of the disks. Any less of a workload wouldn't necessitate 8 GB of cache. Even for a temporary workload this is true; if the disks sit idle for 23 hours a day but for an hour the cache is full (or almost full) and the disks are totally busy, then during this heavy usage time the data is in quite a bit of danger of being lost/corrupted.

      My basic point is; when you need a cache for performance reasons, you are going to be in a situation where the cache has a significant amount of dirty data that would be lost in a crash/power failure. Otherwise you wouldn't need the cache to begin with. Again, some workloads might not suffer a lot of data corruption even under this scenario; storing video or other large contiguous blobs of data will just suffer some loss of a small portion of any given file, which will be relatively obvious. Transactional data is at a much higher risk because the entire database could become unusable because of out-of-order writes (which is the primary method caches use to speed up disk writes) and a crash.

      I also think the trend is to more cached writes, not fewer of them, and often in ways you can't control the latency of the cached writes (think SANs or other remote storage where not having cached writes would mean huge latency penalties and unusable storage).

      Any SAN worth buying will have battery backed write cache, and may even support in-order writes for transactional loads that require it.

  51. Re:You don't want software raid and the (cheapest) by Creepy+Crawler · · Score: 1

    What if it's for a small, say =6 person office? 4 (5400RPM) SATA drives are perfectly fine.. if they're document and spreadsheet workers. The situation is completely different if these were video editors or CAD/CAM software types. Then you need 10GBps server, hardware raid10, say 15 drives, quad-core, and max ram.

    --
  52. Roll your own by Rainwulf · · Score: 1

    Pulling over 70 megabytes a second using a 90 AUD motherboard, with a 80 AUD amd cpu and using silicon image sata cards with 1tb WD green caviar hard drives, and running windows 2003 standard. Now you dont have to run 2003, Solaris would be your next best bet for the beauty that is ZFS, or just run something along the lines of freenas. If the os doesnt like the onboard nic (atheros L1 in my case) you can get a PCI-E intel gigabit desktop card for 70 AUD, which should run just beautifully. Like everone said, if you want speed, dont even bother with the small nas products. Its just not going to happen. In your application you need a real proper server, and a non embedded OS.

  53. Would not recommend the Buffalo by macraig · · Score: 1

    I bought a Buffalo NAS about three years ago; I bought it because of the 1000base-T interface and low cost. I persevered with it for about three months, and then demanded and got a full refund from the retailer.

    1. Re:Would not recommend the Buffalo by Anonymous Coward · · Score: 0

      You "persevered" with it? What the fuck exactly does that mean you did with it? What was wrong that you had to return it?

    2. Re:Would not recommend the Buffalo by macraig · · Score: 1

      Directly pursuant to the topic of the OP: very poor network throughput, as well as discovery just how marginal its design, components and construction was. The deal-breaker was the poor data-rate, because that's why I had the gigabit network set up in the first place, when I could have otherwise settled for cheaper 100base-T components; that wireless/wired gigabit router set me back a bit at that time, when equivalent 100base-T ones were plentiful and dime a dozen.

      I wound up building a RAID 5 subsystem internal to my main system, based around an LSI MegaRAID HBA. I got 40% more storage and full drive bandwidth for the same cost as the Buffalo.

  54. tiny cube pc,few drives, and openfiler or solaris by itzdandy · · Score: 1

    get a small pc case such as one of the many small cube cases that come as a barebones. Put a dual core chip and 2GB ram. The you can install something like openfiler which will give you a nice web interface and the ability to do nfs,cifs,ftp,and iscsi. Alternatively, install solaris or opensolaris and use ZFS and have the ability to compress the files at the filesystem level and also do a raidz with 3 drives for reliability and speed.

    either way you can bond two ethernet interfaces together for 2Gbit which should get you 80-100MB/s realworld bandwidth.

  55. Check out this brief synopsis of the ReadyNAS-pro by Wild_dog! · · Score: 1
  56. Features vs. speed by eagl · · Score: 1

    I have an Infrant ReadyNAS+ and it is not fast. It has a TON of features (most of which I don't use) but transfer speeds are pegged at approx 7% to 8% network utilization through a gigE switch even with jumbo frames on and an upgraded stick of ram for the NAS cache. I get the same transfer rates with 3 different computers of various types including an older laptop and a very fast gaming machine, and my transfer rates are fairly close to what others report, which tells me the bottleneck is the NAS device. There may have been some improvements in the device since infrant sold the product line, but you'll need to check their support forums to see what people are reporting with the newer ones.

    If you don't need many of the easy to use features of most of the low-end NAS devices, you are probably better off rolling your own. Even using the cheap embedded raid chips on consumer mobos and what you would probably consider a bottom-end cpu (like a single core celeron) is going to get you faster transfer rates than many of the NAS devices on the market. There are a few ready to go home router linux distributions that ought to be fairly secure and feature-rich, and they will probably grossly outperform a consumer level NAS box. The only tradeoff will probably be power
    consumption, however if you pick decent components, allow the cpu to throttle down, and let the drives halt when not in use, you can minimize the difference in power usage to the point where it might cost an extra $50ish/year over a purpose-built NAS device.

  57. Try an appliance like this. by Anonymous Coward · · Score: 0

    Try this unit.

    http://www.acnc.com/02_01_jetstor_sata_405u.html

  58. Build a shuttle box by bummster · · Score: 1

    Sounds like you want and maybe need a turn key solution that is cheap. But as they say, Fast, Cheap, Reliable... Buffalo LinkStation Quad could be an appealing solution. It gives you Time Machine for your macs, but is probably a dog with datasets larger than 1GB. At a 1GB file size, smallnetbuilder.com shows write speeds to be 11.4MB/sec. If and when you do ultimately decide that soho solutions can't give you the speed you need, be sure to load the pc with as much RAM as it can take. Ample RAM to cache the file system can either maintain or reap huge throughput gains.

  59. Build yer own by Gothmolly · · Score: 1

    I have a Linksys NSLU2 (running Debian Lenny) and it maxes out at about 4 MB/sec, the dinky amount of RAM means almost no FS or other buffering is possible, and the limp CPU (266 MHz) just can't push the IO fast enough.
    A barebones PC is probably $200 + drives, slap OpenFiler or a real distro on it, and share out 1 big /share filesystem via Samba.

    --
    I want to delete my account but Slashdot doesn't allow it.
  60. Best SoHo NAS in my opinion by Anonymous Coward · · Score: 0

    QNAP 409-T.

    Never let me down. Highly recommended.

  61. OP: "off the shelf" by StaticEngine · · Score: 4, Insightful

    See, the problem with responses like this is that they ignore the request of the original poster, and, while being valid instructions for a home-built, it is only a good solution if the time of the OP has zero value. Your instructions involve eight steps: Order (multiple) parts, wait for delivery, assemble, learn how and then install OS, learn now and install three other packages. The OP is looking for three steps: Order one thing, wait for delivery, plug in and use.

    Your post has value to the DIY crowd, certainly. But for someone looking for a product recommendation, it totally missed the boat.

    1. Re:OP: "off the shelf" by ceoyoyo · · Score: 1

      True, but the submitter also specified that he wanted it fast (most important criteria) and he specified SOHO so I assume there's a budget.

      If you want speed on a budget your best bet is going to be to use a PC. You can order it pre-built if you want, but you might have to do a bit of configuration. The other option is to buy a non-SOHO NAS for a lot more money, or make due with the underpowered ones you can get for home use.

    2. Re:OP: "off the shelf" by PiSkyHi · · Score: 2, Informative

      There's a paradox in it - those who know how to build a NAS would never buy a ready made 1.

      That's also the reason ready made NAS's don't have the features required, because only the people who don't know what to look for, buy them.

    3. Re:OP: "off the shelf" by nabsltd · · Score: 1

      The best bet, then, is a pre-built PC of decent specs (dual or quad core CPU) with room for plenty of SATA drives.

      You don't care about RAM installed, only about the maximum the system supports, because you can then go and buy RAM to fill the machine for far less than Dell, HP, etc., charge.

      The theory is the same for hard disks. You just need 6-8 SATA ports on the motherboard, and then you can add as many drives as you want, and use software RAID (or ZFS) to set up the storage.

      If the system doesn't have a decent Ethernet card, a server-class dual-port PCIe x4 card will set you back less than $140.

      Going with a low-end server-class system might be slightly better, but a quality workstation PC with 2-3 year warranty would almost certainly be good enough.

      I've built systems exactly like the story poster wants using inexpensive parts. My storage server has four 1000Mbit Ethernet ports bonded (using Linux bond driver) and can sustain about 300MB/sec from the various systems that acesss it regularly (two ESX servers and a Windows 2003 server) via iSCSI. Plus, it's still got a lot of room to grow (only 2GB of RAM out of 8GB max, and 500GB drives instead of the 1.5TB you can buy now), which this is another advantage to a true server. Generally, you can upgrade it when your needs increase, while a NAS device will probably require you to buy a new one.

    4. Re:OP: "off the shelf" by Anonymous Coward · · Score: 0

      And I respectfully disagree for one reason: it's a vote that there are no good "off the shelf" implementations, or that an "off the shelf" implementation doesn't match your needs. Whether or not you agree with that is a different story, but there are certainly application domains where that -is- true, and the only answer for a "good" solution, or one that matches the criteria, is to roll your own.

    5. Re:OP: "off the shelf" by Kent+Recal · · Score: 1

      Well, depends on what is more valuable, his time or his money. He said SOHO, thus many people assumed his budget will probably not stretch into the netapp/EMC realm.
      Thing is, for the price of real plug'n'pray NAS that will actually push at GigE speeds you can easily hire someone to assemble a box for you - and still save money.

      If you have the dough and really don't care then don't ask slashdot. Call up netapp, they'll be glad to sell you a v3000 or similar box in a shiny package. For a premium they'll also send someone to plug it in for you.

    6. Re:OP: "off the shelf" by BlackCreek · · Score: 1
      I could have built a NAS, but choose (?chose?) to buy a ReadyNAS Duo. While there are a million inconveniences to it, there are the following unbeatable points:

      1. SIZE;
      2. Time of my life it cost me to have a NAS running at home.
      3. Assurances regarding noise, and power consumption.

      I probably could have researched enough to be sure of the noise level of whatever I would buy, but it would have cost way too much time. In any case, I doubt I could have built something as small.

    7. Re:OP: "off the shelf" by PiSkyHi · · Score: 1

      I want small in something I can take with me. I want quiet in something that has to be near me.

      With any hardware purchased losing value constantly, I also want the best value.

      I just can't bring myself to pay for something I don't need. I do need something, somewhere safe to keep my files though.

    8. Re:OP: "off the shelf" by jp10558 · · Score: 1

      For me, at home, the problem was storage space. If it only holds 2 disks, why wouldn't I just drop them in my tower and share them out from my workstation?

      I want at least 6 disk and dual stripe parity...

      --
      Opera, Proxomitron-Grypen,GPG 0x0A1C6EE3
  62. Already been extensively discussed... by kwabbles · · Score: 5, Informative

    For example:

    Best home network NAS?
    http://ask.slashdot.org/article.pl?sid=07/11/21/141244&from=rss

    What NAS to buy?
    http://ask.slashdot.org/article.pl?sid=08/06/30/1411229

    Building a Fully Encrypted NAS On OpenBSD
    http://hardware.slashdot.org/article.pl?sid=07/07/16/002203

    Does ZFS Obsolete Expensive NAS/SANs?
    http://ask.slashdot.org/article.pl?sid=07/05/30/0135218

    What the hell? Is this the new quarterly NAS discussion?

    --
    Just disrupt the deflector shield with a tachyon burst.
    1. Re:Already been extensively discussed... by Anonymous Coward · · Score: 0

      good thing that nothing new or useful was built or coded between 2007 and today...

    2. Re:Already been extensively discussed... by Anonymous Coward · · Score: 0

      Yes, slashdot must periodically invent these to stir up some reader excitement. "I'm brand new on the job, and my boss just asked me to design a {$10 million machine room, supercomputer, next-generation content distribution network, new programming language, plan to outsource all our jobs to the third world.} Does anyone here have any advice?

    3. Re:Already been extensively discussed... by sootman · · Score: 3, Insightful

      What the hell? Is this the new quarterly NAS discussion?

      Yes, I hope it is. Maybe not quarterly, but I have no problem "revisiting the classics" periodically. Technology marches on, best practices come and go, so it is useful to cover the same ground every so often. Seven years ago the coolest story ever was covered here: build a Terabyte fileserver for less than $5,000!!! (Note to visitors from the future: it is late 2008 and you can buy an external terabyte hard drive for a little over $100. Call it $125. That same five grand could buy you FORTY terabytes today. You probably got a 1TB USB jump drive in your cereal this morning.)

      Plus, not everyone has been around as long as you and I. Won't somebody please think of the n00bs?!? :-)

      --
      Dear Slashdot: next time you want to mess with the site, add a rich-text editor for comments.
    4. Re:Already been extensively discussed... by WuphonsReach · · Score: 1

      Eh, for $5000, the upper limit is probably around 15-spindles. Including the cost of the case ($600), a good motherboard and CPU and RAM ($600), a good 16-port SATA controller ($1000), and all of the drives (15 drives @ $130 each adds up quick). Okay, well, that's only about $4000.

      Figure RAID-6, 1 hot spare, net useful space of (12) 1.5TB disks. Which is about 18TB (probably closer to 16.5TB once you factor in over-head and the million bytes vs megabytes conversion).

      Still not shabby at all.

      --
      Wolde you bothe eate your cake, and have your cake?
  63. gbe performance by Anonymous Coward · · Score: 0

    i have written the drivers for several
    nas products. i have benchmarked several
    nics (pci and pcie) at 125MB/s (1gbit).
    the key is jumbo frames. even 4k jumbos
    will do.

  64. I just bought the 4t WD... check it out by dmreference · · Score: 1

    WD ShareSpace(TM) Network Storage System - 4 TB, Gigabit Ethernet http://www.wdc.com/en/products/Products.asp?DriveID=501 I found it for a little over $800. I went back and forth with several options the Drobo and MS Home Server included. I came to conclusion that, Home Server does not provide the level of fault-tolerance I wanted in a solution so that was out. The Drobo is cool and all, but I don't like the extra $200 for the networking. Since I was going to go buy WD drives anyway, why not get the whole thing from them ;)

    1. Re:I just bought the 4t WD... check it out by martinX · · Score: 1

      "Remote access software - Access your files anywhere, anytime using MioNet® remote access services from WD. "

      I don't like the sound of requiring a third-party 'service' to access files over the internet. Do they still block certain filetypes from being transferred?
      http://www.pcadvisor.co.uk/news/index.cfm?newsid=11576

      Also, I checked out the five reviews WD linked to. Of those 5, 3 were identical. So of the three actual reviews, a throughput of about 10 MB/sec was the norm. Not exactly Gigabit speeds.

      This NAS buying thing is harder than it looks :-/

      --
      When they came for the communists, I said "He's next door. Take him away. Goddam commies."
    2. Re:I just bought the 4t WD... check it out by bertvv · · Score: 2, Insightful

      What misses in the specs is which processor is inside. I've got a MyBook World Edition that also has GigE, but the processor is so underpowered it barely reaches 10MB/s

  65. Good Review Site! by Anonymous Coward · · Score: 0

    You should check out www.smallnetbuilder.com. Lots of great reviews of SoHo type NASes. I've been happy with my Thecus N4100PRO, built around the AMD Geode processor.

  66. Understand your performance requirements by raw-sewage · · Score: 3, Interesting

    How many gigabytes are "multiple" gigabytes? Seriously, moving around five GB is much easier than 50 GB and enormously easier than 500 GB.

    Another thing to consider: how many consumers are there? A "consumer" is any process that requests the data. If this post is a disguised version of "how do I serve all my DVD rips to all the computers in my house" then you probably won't ever have too many consumers to worry about. On the other hand, I work for an algorithmic trading company; we store enormous data sets (real-time market data) that range anywhere from a few hundred MB to upwards of 20 GB per day. The problem is that the traders are constantly doing analysis, so they may kick off hundreds of programs that each read several files at a time (in parallel via threads).

    From what I've gathered, when such a high volume of data is requested from a network store, the problem isn't the network, it's the disks themselves. I.e., with a single sequential transfer, it's quite easy to max out your network connection: disk I/O will almost always be faster. But with multiple concurrent reads, the disks can't keep up. And note that this problem is compounded when using something like RAID5 or RAID6, because not only does your data have to be read, but the parity info as well.

    So the object is to actually get many smaller disks, as opposed to fewer huge disks. The idea is to get the highest number of spindles as possible.

    If, however, your needs are more modest (e.g. serving DVD rips to your household), then it's pretty easy (and IMO fun) to build your own NAS. Just get:

    • a case that can hold a lot of disks
    • a fairly recent motherboard
    • the cheapest CPU supported by the motherboard (your load is virtually all I/O; very little CPU is needed with modern I/O chipsets)
    • some RAM
    • a high quality, high capacity power supply
    • the disks themselves
    • and your favorite free operating system of choice

    You might also want to purse the Ars Technica Forums. I've seen a number of informative NAS-related threads there.

    One more note: lots of people jump immediately to the high performance, and high cost RAID controllers. I personally prefer Linux software RAID. I've had no problems with the software itself; my only problem is getting enough SATA ports. It's hard to find a non-server grade (i.e. cheap commodity) motherboard with more than six or eight SATA ports. It's even harder to find non-PCI SATA add-on cards. You don't want SATA on your PCI bus; maybe one disk is fine, but that bus is simply too slow for multiple modern SATA drives. It's not too hard to find two port PCI express SATA cards; but if you want to run a lot of disks, two ports/card isn't useful. I've only seen a couple of four-port non-RAID PCIe SATA cards. There's one eight port gem, but it requires PCI-X, which, again, is hard to find on non-server grade boards.

    1. Re:Understand your performance requirements by OneMadMuppet · · Score: 1

      The Digitus DS-30104 is a 4 SATA PCIe card for about 50. I've not tried it though, I don't know which chipset it uses and I don't know if there is *nix support... ...but if anyone else does - it would be great to know ))

    2. Re:Understand your performance requirements by mistahkurtz · · Score: 1

      requires PCI-X, which, again, is hard to find on non-server grade boards.

      and is going away in favor of pci-e.

      --
      not only is time travel possible, it's irrelevant.
  67. Have you looked at Delta-V ? by Anonymous Coward · · Score: 0

    These guys make fast storage units. Delta-V was announced last week, and they showed it off at SC08 a few weeks ago. I saw their 3U unit in some booth. It was attached to a Windows machine as an iSCSI target, and the guy running the demo was the company founder, and had IOmeter running on it. He got 500 MB/s across the iSCSI link. He said it hooks to anything, windows, linux, Mac, and unix.

  68. I like the Synology boxes (CS508 or RS407) by crispytwo · · Score: 2, Informative

    They don't do too badly for xfer speed and are quite reliable. They seem to use less power and aren't noisy like other NAS systems (especially the RYO).

    Linux is their OS and if you need to add some functionality, you can get in and do it, but it works well out of the box.

    RAID 5 or 6 with the 508

    I've done the Windows SMB and it sucks for maintenance and you're back at RYO - patch and crotch rub. I've built many a linux box for this and, though they work, I have better things to do with my time. I really appreciate buying a few HD and sticking them into a box and having a system that can store data, xfer data, backup themselves, etc. in a matter of minutes.

    Oh yes, compatible... via CIFS with most systems. NFS with Mac and Linux if you are so inclined. rsync for backup.

  69. none by Anonymous Coward · · Score: 0

    If you're that concerned about speed, you might want SSD drives instead, seeing as the network transfer speed will appear slower due to conventional hard disk write speeds.

  70. Solaris and ZFS by linuxtelephony · · Score: 1

    Consider Solaris + ZFS too. Especially now that Solaris 10 u6(?) now can install to ZFS root partition (HINT: Use Text installer - options 3 or 4 if memory serves).

    Solaris is free as in beer, even if it isn't open source. Plus you get the benefit of some of the proprietary drives if you have older hardware. Plus, Solaris proper won't leave you in a lurch when things change in OpenSolaris and you can't do updates or run some programs. [Admittedly this problem seems to be mostly resolved, but for mostly production environment I'd suggest Solaris over OpenSolaris unless you need some particular bleeding edge feature not yet migrated from OpenSolaris into Solaris.]

    I did just this a while back when looking for a storage solution for backups. The SOHO options did not have the bandwidth, even with gigabit nic ports. In the end, moving to PC hardware with SATA drives worked much better.

    Due to SATA controller issues with port multipliers when I set this up a year or two ago I ended up having to switch to Linux with md. Regardless, the performance difference was dramatic and the PC based system actually worked quite well.

    --
    . 62,400 repetitions make one truth -- Brave New World, Aldous Huxley
    1. Re:Solaris and ZFS by UU7 · · Score: 1

      Opensolaris is much easier for first times to set up, check out http://breden.org.uk/2008/03/02/a-home-fileserver-using-zfs/ for a guide that I found helpful.

    2. Re:Solaris and ZFS by myz24 · · Score: 1

      Solaris isn't free, it is a 90 day trial. I went to setup a file server using Solaris and quit after reading the fine print.

      I tried OpenSolaris but it wouldn't recognize any of the drives in the system.

    3. Re:Solaris and ZFS by pyite · · Score: 1

      Solaris isn't free, it is a 90 day trial. I went to setup a file server using Solaris and quit after reading the fine print.

      Uh. Complete and utter FUD. From the horse's mouth:

      You can use the Solaris 10 Operating System at home or at work -- without paying a license fee. Download the Solaris 10 OS today - FREE -- when you register your system (as part of the download or by going through the registration portion of the download) you will receive an Entitlement Document, which grants you unlimited rights to use Solaris registered machines. Sun offers this no-cost licensing program for all use - even commercial use, even on multi-processor machines.

      --

      "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

  71. LaCie 1 Tb by Anonymous Coward · · Score: 0

    It canhaz the standard RJ45 jack. Into the hub it plugs.

    Bought this a year ago from PCConnection (post raccoons). Plug into the network and store. Works with Mac Win Lin. $300 on sale.

    Just do it.

    Oh, and I run at work a network with Leopard, Tiger, DOS3.3 Q&ADOS, Win98, WinXP, MintLinux, SAMBA, and the boss's stupid kids scratch the last one. Don't recommend Clarknet, stick with rolling a MintLinux.

  72. Synology boxes are a GREAT by gelfling · · Score: 2, Informative

    They are a little on the high end, cost wise for consumer boxes but they are very reliable, the firmware actually works WELL, they support NTFS and their network interfaces function up to spec. And they support Mac.

    They make units from 1 bay SATA up to 4 bay 1U hot swappable dual 1Gb dual power supply rackmounts.

    www.synology.com

  73. along the lines of sneaker net by fred+fleenblat · · Score: 1

    I'm going to suggest that you skip the NAS and just get a large-capacity eSata or firewire drive. Plug it into your current test machine, do your thing, unplug it and move along to the next machine. This approach sidesteps any limitations of your LAN, host machine, RAID cards, or NICs.

  74. Jumbo Frames, trunking, TOE cards by Luminary+Crush · · Score: 1

    You might want to investigate running a data network w/ jumbo frames (change ethernet MTU size from 1500 to 9000).

    1500 MTU was good when networks were 10M, but with Gig they are really just too small. The CPU gets pummelled with packet assembly/disassembly; setting the MTU to the jumbo (9000) size reduces the required CPU overhead greatly. However, all devices on the network segment must be set to jumbo (9000) byte MTU.

    On a larger corporate network we'd setup a private, non-routed VLAN for NAS; all servers involved would have a second NIC attached to this VLAN at 9000 MTU. You could still manage this in a SOHO with multiple NICs on a separate LAN/switch (be sure the switch supports jumbo frames) just for storage. These days a good percentage of motherboards come with dual NICs anyway.

    Jumbo frames can significantly increase ethernet throughput, particularly with large streaming I/O, which sounds like what you have.

    Another option, particularly on your NAS/server, would be to setup an ethernet trunk (also called 'etherchannel' or 'nic teaming') - logically bonding two network interfaces to send/receive data as one IP address. If you have multiple clients uploading/downloading data to the server at the same time this is essential.

    All 'modern' operating systems support jumbo frames and trunking; many of the cheapo SOHO ethernet switches do not, so you might have to shop around. I think I've seen models which support both jumbo frames and trunking for under $300.

    You could also investigate getting TOE (TCP Offload Engine) cards which have 'packet co-processors' to perform the packet assembly/disassembly of the small 1500 byte packets - but these are usually pricey and jumbo frames often work just as well.

  75. DIY by Efialtis · · Score: 1

    I researched Drobo, and found it to waste too much space, not to mention that the network add-on is expensive...and then reduced to usb transfer rates... I have a Buffalo on my desk, and it is ok, but if you need speed, that isn't a way to go. I found that I could buy a real RAID card and roll my own RAID NAS with little trouble and get great speeds out of it. Later I purchased a micro board, it has 6 sata ports, and I have a Software Raid running on Ubuntu server, with all the necessary bits for SAMBA, XFS, CIFS, etc. I have 6 disks all in raid and a Gigabit ethernet...faster than anything I have purchased previously...does a good job too...

    --
    --E--
  76. TheCus make higher performing NASes by belorion · · Score: 1

    I have several thecus NASes (http://www.thecus.com/) the N5200, the N5200Pro and I'm looking into a N7700 or a N8800. They all have Gigabit ethernet connections. They are more expensive than purpose built boxes - but they are "set it and forget it" appliances - which is where there value comes from.

    I can sustain approx 25MB/s write and 40 MB/s on my N5200. NOTE: That's bytes not bits. In otherwords 200Mbit read throughput and 320Mbit write throughput. This is far better than the other domestic of SOHO NASes - which is why thecus tends to win awards http://www.thecus.com/news_contentx.php?nid=827 , http://www.tomshardware.com/reviews/holiday-gift-guide,2065-12.html

    If you contrast this with one of my windows RAID arrays (100+MB/s read 100/MB write - array to array or just streaming) it looks rather poor.

    If you want something that you can just forget about and will give reasonable performance then TheCus is what you want ... if you want better performance than a custom built server is what you want (Windows or Linux).

    1. Re:TheCus make higher performing NASes by belorion · · Score: 1

      Ooops. That should state 25MB/s write and 40 MB/s read ... and 200 MBit write and 320 MBit read

  77. BuffaloTech sucks. by Anonymous Coward · · Score: 1, Informative

    I had their TeraStation a few years ago. I bought it from Newegg, whose site (at the time) said that the TeraStation came with a 2 year warranty. 1 year after I bought it, it started acting funny. I called up BuffaloTech, only to be informed (after a near 2 hour wait on hold), that the TeraStation warranty is in fact only ONE YEAR and that they DO NOT repair TeraStations out of warranty. Yes, you heard that right... they won't even repair it and bill you. The jackass had the audacity to tell me that I should buy another one to get my data off. I told him to fuck off. I plugged the hard drives into my linux box and got the data off myself. Assholes.

    P.S.--Newegg saved the day. At first, they told me to go fly a kite. After asking them very nicely to ask their manager, they said OK and issued me a RMA number. I got my money back minus a small restocking fee (which is reasonable considering I didn't have the original box anymore). Newegg FTW.

  78. clarkconnect.com by Anonymous Coward · · Score: 0

    That's what I meant. Gawd, it's weird. Just roll the standard Samba.

  79. any write? by kentsin · · Score: 1

    You say test, do they change the data set?

  80. Professional Networking by ZP_558963 · · Score: 1

    I do not know if you have automated anything yet because from your description it sounds like it is still a very manual system. The Linux box I am assuming is a test box. Also it sounds like you are looking for a centralized storage location. It also sounds like this is using a mixed wired and wireless system. In my opinion, there has been very little thought put into how your network and future expansion.

    For speed which it sounds like is your immediate concern, Current OS-X boxes have Gigabyte Ethernet capability. Most likely you have a single Gateway for the entire office. If this is the case, an upgrade is highly recommended. I would also look into your Windows and Linux box network cards. Default hardware by Dell and other major distributors tends to be cheap equipment. Cheaper gateways will reduce the speed of the network to the slowest entity. Which is your wireless device. Having separate Gateway and wireless devices may be something else to look into. A more advanced Gateway/Router will be able to allow you to easily change speed.

    As far as a networked hard drive and Lan gateway, please don't do that. You are playing with fire. It is good for consumers for light duty use, but companies that need throughput and speed, it will die within the year. Many people that posted to create your own Linux storage box are correct. A single box is easily expanded into a larger server. You can buy an off the shelf system. I would also suggest that you do not use a test machine. Also you can create a Raid disk array to make swapping drives that die as painless as possible.

    I did this professionally for many years. To conclude, upgrade your gateway and add a new computer for data storage.

  81. Yes it will (well, crap SOHO cpu/network). by thesupraman · · Score: 3, Insightful

    Ah, wrong.

    This guy is talking about SOHO type NAS boxes, their cpu and network throughput is their bottleneck.

    If he was talking about 'real' NAS, then that is very different (although it is still trivial to get a NAS that can saturate GBit for many workloads).

    Our 16/32 drive Raid6 SATA raid arrays easily sustain 400MB/sec locally for moderately non-random workloads - there are workloads for which this of course does not apply, but since he is apparently moving around GByte lumps, it would not be his case.

    SOHO NAS devices normally run out of grunt at around 6MB/secish, even for long linear reads, some do better at up to 25.

    I am thinking your workload is TPC type database loads, dont assume everyones is (we have a mix of video files and software development, very different..). TPC type disk loads are a corner case.

    We also love ATAOE but that is DEFINITELY not what he is looking for.

  82. Re:UnRaid: when build-from-scratch isn't fast enou by Enigma2175 · · Score: 1

    Okay, unRaid is not particularly fast compared to an optimized system, but it's expandable, had redundancy, is expandable

    Good thing your post has redundancy as well, otherwise we wouldn't know it was expandable.

    --

    Enigma

  83. Nope... by Anonymous Coward · · Score: 0

    im not helping you and your warez
    thats what killed the economy

  84. DROBO by randomsample · · Score: 1

    I also work for a smaller company and I have 2 Drobos in service for backups. They are working fine but you are limited to USB 2.0 speed. This is true when you use the associated DroboShare network appliance as the connection from the Droboshare to the Drobo is USB. BackupExec does not like the DroboShare so I have them directly connected to a server. This arrangement keeps us from running all that data across the network as well. The only downside to the Drobo (this is unsubstantiated as I have not tested it) is they are supposed to be susceptable to data loss resulting from loss of power. If you are operating a data center with adequate UPS and a backup generator this would not apply. 4TB of redundant SAN for less than 1K is hard to beat, we'll see how long they last.

    --
    Of all the things I've lost, I miss my mind the most. Or do I?
  85. NAS Charts by Anonymous Coward · · Score: 0

    http://www.smallnetbuilder.com/component/option,com_nas/Itemid,190/

    "This test measures the average filesystem performance of NASes at the Ethernet connection speed"

  86. ReadyNAS Pro 3TB Version by Anonymous Coward · · Score: 0

    I have the Netgear ReadyNAS pro. It has a intel core duo processor running Linux on it. A fairly nice interface for managing the device from the web browser. Plus if you want (and I did), you can install a module provided by Netgear to get root access to the box. You risk voiding your warranty depending on what you do, but you can then tweak the box to do more, etc.

    Performance has been good. I've got a 5 port Netgear switch (1Gb). From my MacBook pro 2.4ghz via AFP, I can get a sustained write of 41 megabyte / sec. I don't know if it can go faster or not writing, as that might be my laptops read max. Reading from it, I was only able to get 39 megabyte / sec...but that again might be the max write speed of my drive...?

    My ReadyNAS pro has three 1TB drives in it. There was also a rebate on it for a $400 video camera making the incentive to get it a little better.

    I've been very happy with the speeds of it.

    --Ben

  87. One of these? by rthille · · Score: 1

    http://sfbay.craigslist.org/pen/sys/955430088.html

    Dual-Core Intel box with 12 SATA drives for 3TB total storage. 3Ware 12-port raid card.

    $1000

    Since it's for a business, it probably makes sense. For my home, too loud, big & power hungry.

    The QNAP TS-509 looks interesting for my home network, but it's expensive and I'm unsure about the software....

    --
    Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
  88. The BEST answer is by jhines · · Score: 1

    The computer your currently using, then you can get a better one!

    Solaris and ZFS do wonders making PC's into NAS servers.

  89. I want a cheap SOHO NAS that acts like a real one by EmbeddedJanitor · · Score: 1
    That's what he's really saying. He's hoping to get a grunty NAS at Computas-R-Us prices.

    Sorry folks: if you want grunty corporate grade equipment with high performance then don't expect to buy it at a retail shop.

    Build your own or spend up on professional grade equipment.

    --
    Engineering is the art of compromise.
  90. Skip Software RAID boxes by aaarrrgggh · · Score: 3, Informative

    The Buffalo Terastation uses a software RAID, which slows it considerably, with the side benefit of being nearly impossible to recover if it crashes.

    It does support SMB, NFS, and AFS out of the box though.

    These boxes are cheap crap, and have a very limited useful lifespan. Our company lost a good deal of information when ours crapped out after 366 days. (Yes, we had backups, No they weren't perfect. They happened to be with me halfway around the globe at the time...)

    Really seems like the product offerings in this space are limited usability, poor reliability, imperfect implementations, and grossly overpriced. Doing it over again, I would go for a build-it-yourself box hands down.

    1. Re:Skip Software RAID boxes by TCM · · Score: 1

      Software RAID (as in the OS manages the disks and has its own device layer doing the RAID) is actually superior to anything else in terms of flexibility and recoverability.

      Got hardware involved doing your RAID parity computation? Good luck getting a replacement card some years down the road, since most have their own proprietary way of storing the signature that says "this disk is 1 of 5 of a RAID5".

      And pure software RAID being slow is a thing of the past or a myth of people setting up their RAIDs wrong. >200MB/s reads and writes with barely any CPU load fast enough?

      --
      Of course it runs NetBSD. BTC: 1NT7QvbetmANwaMzhpVL6
    2. Re:Skip Software RAID boxes by swillden · · Score: 1

      And pure software RAID being slow is a thing of the past

      Exactly.

      In fact, if you do some benchmarking you'll find that Linux software RAID is MUCH FASTER than most hardware RAID cards if you have a decent CPU that isn't loaded down with other computation.

      Where hardware RAID shines is when your CPU is running flat out on the computations that the I/O is serving up. In that case it obviously helps to push the parity computations off on another processor, the one in the RAID card. However, if your workload is not CPU-intensive, a typical dual-core 2.5-3 GHz processor will wipe the floor with all but the highest-end hardware RAID cards.

      Software RAID is actually superior to anything else in terms of flexibility and recoverability.

      And then there's that. I've done some crazy stuff with Linux software RAID that no hardware card would ever do, and I've recovered from at least one situation that any card would simply have given up on. For that matter, Linux MD didn't want to touch it, either, but the --force and --assume-clean options can be your friends in very dark times.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
    3. Re:Skip Software RAID boxes by Anonymous Coward · · Score: 0

      I've done the software route, and lost information. Still looking for someone, anyone to try and put it back together. Long story.

      Hardware raid hardware raid hardware raid. I learned the hardway.

  91. Thecus 5200b Pro by Mobsta · · Score: 1

    Thecus 5200b Pro - currently the best soho NAS i can find. I use it with Raid 5 and it never drops below 30MB/s throughput. Highly recommended!

  92. Re:UnRaid: when build-from-scratch isn't fast enou by dbug78 · · Score: 1

    and it's also expandable

  93. Drobo + OSX Server by Anonymous Coward · · Score: 0

    Take an older Mac, spend some $$ on a Drobo and OSX Server and you have a easily managed, robust server with redundant storage. If you have the FW800 Drobo, your throughput will be more than typical gigabit can handle. I have this setup on an old G4 PowerBook with OSX 10.5 server. Bonus: With OS X Server, all attached drives can be enabled for Time Machine backups. Way better and faster than a Time Capsule. (Too bad it is 5x the price...)

  94. Forget SOHO boxes by TopSpin · · Score: 2, Informative

    What you're expecting is really beyond the capability of common SOHO NAS equipment. These devices lack the RAM and CPU to approach the capacity of GB Ethernet.

    Unless you're willing to roll your own, you should consider a better class of gear and spend your time arguing for the funds to pay for it (a NetApp S550, perhaps.) If you are willing to roll your own, you can get there for $1-2k using all new hardware.

    Beware reusing older hardware; many GB NICs can't approach GBE saturation, either due to PCI bus contention or low end, low cost implementation. Yes, in some cases older hardware can get there, but this will require careful configuration and tuning.

    You want a PCI-E bus, a decent 'server' class NIC, recent SATA disks, a modern CPU (practically any C2D is sufficient) and enough RAM (2-4 GB). Personally I stick to Intel based MB chipsets and limit myself to the SATA ports provided by Intel (as opposed to the third party provided by jaton, silcon image, et al.) Linux, md raid 10. Will saturate a GBE port all day long, provided your switch can handle it...

    You're serving desktops so jumbo frames are probably impractical (because some legacy hardware on that LAN will not tolerate it.) If your managed (?) switch can provide VLANs you can multihome your critical workstations and use jumbo frames. This will get you more performance with less CPU load for 'free'.

    --
    Lurking at the bottom of the gravity well, getting old
    1. Re:Forget SOHO boxes by Anonymous Coward · · Score: 0

      A NetApp S550? Are you kidding me? To serve up data for a few PCs? That's like telling him to guy buy the $6000 Adobe Creative Suite 4 because he occasionally has to increase brightness on some images.

    2. Re:Forget SOHO boxes by mistahkurtz · · Score: 1

      What you're expecting is really beyond the capability of common SOHO NAS equipment.

      I just re-read the OP's request, and I remembered a client of mine, a little over a year ago, in a small office, around 30 people. Spent around $15k for an entry-level SAN from HP, MSA 1500i. i guess my question for the OP would be, what do you mean by SOHO? Are you strictly speaking price? And, if so, what do you consider SOHO prices? If you're looking to not spend $40k for an equallogic/lefthand/etc SAN, but are ok spending a small chunk, then maybe an entry level iSCSI SAN from HP, EMC, etc will work. I can't speak to every make and model, but most of these are built around "server quality" hardware, typically reasonable RAM, Xeon-class CPUs, dedicated network and drive controllers, etc. HP and some of the other OEMs also sell "storage servers", for HP this would be based off of a DL380 chassis, populated with disks, running windows storage server.

      i guess what would help me understand your request would be more details. what do you consider SOHO? what is a reasonable price to pay to get what you want?

      --
      not only is time travel possible, it's irrelevant.
  95. Trying to re-ask the question... by Anonymous Coward · · Score: 0

    I'm going to go out on a limb and say that the original poster might not be thrilled with the responses... which are the same sort of responses that usually get posted in response to a question like this.

    I'd like to ask the exact same question again, while adding the additional condition that I ABSOLUTELY DO NOT UNDER ANY CIRCUMSTANCES WANT TO BUILD AND MAINTAIN YET ANOTHER SERVER BOX TO DO THIS!!!!!

    Yes, I know it's probably cheaper, faster, and (most importantly) geekier, but at the moment, I'm already up to my big fat overworked ass in machines that need to be managed, both at work and at home, and what *I* need at home is a little box full of hard drives that I can plug in and share files across the Linux, Macos, and Windows platforms with WITH AS CLOSE TO ZERO ADMINISTRATIVE OVERHEAD AS IS FEASIBLE.

    I have a little DLink DNS-323, but it kind of sucks; it is very slow, cannot effectively use my gigabit home network, and I do not trust it.

    I've heard good things about the Infrant... er, now Netgear, I guess? I've heard a few good things about Drobo. I've never used the Buffalo stuff, but I bought a couple of their external hard drives a couple years back, and they were crap, so I'm disinclined to buy a higher-end item with them.

    I have tried FreeNAS, and for what it is, it's actually fairly nice, and for those inclined, I'd recommend investigating it, particularly if they've finished the administrative interface for groups and stuff, which they hadn't when I last tried it. But, to reiterate, that sort of solution is NOT WHAT I AM LOOKING FOR!

    So, can anyone recommend a little toaster-size box that sits quietly and shares files, and doesn't require me to build the hardware and then install software? Hmm?

    We've installed a couple of the Infrant/Netgear ReadyNAS boxes at work for various people, and haven't had many complaints. So I'm guessing that's what I'll go with, someday. They're just a little pricey; and I don't know what the performace is like.

    1. Re:Trying to re-ask the question... by rthille · · Score: 1

      What's your price range? Minimum storage capability? The QNAP TS-509 is ~$900+drives, holds 5 drives, looks reasonably fast, has online RAID migration, and an open OS (linux) with ssh shell access and package support.

      I'd probably buy one if I hadn't invested in a new server last year.

      --
      Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
  96. One more thing ---- Version control by Anonymous Coward · · Score: 0

    Since you are probably not creating umpteen gigs of data every day, would it be possible to subdivide the data into smaller managable files and use version control to sync things up?

    CVS over NFS is pretty fast. Subversion might be better for binary files.

  97. This is misleading. by Jane+Q.+Public · · Score: 3, Interesting

    While it is true that the outside of the disk is spinning faster than the inner portion, in a modern HDD there are also several times more sectors in those outer rings. So while strictly speaking the read times might be faster, the seek times are not, and may even be slower. The sectors might even be interleaved, making any such comparison almost meaningless.

    However, as you say, benchmarking is the only way to really tell. Highly recommended.

    1. Re:This is misleading. by cyphercell · · Score: 1

      "The sectors might even be interleaved, making any such comparison almost meaningless. "

      Also misleading, you can compensate by partitioning the disks and use the faster partitions before you use the slower ones. I believe this is often done for benchmarks etc.

      --
      Under the influence of Post-Cyberpunk Gonzo Journalism
    2. Re:This is misleading. by Jane+Q.+Public · · Score: 1

      That's what I wrote. Please read again.

      All I meant was that simply saying that the outside of the disk spins faster is not grounds for using that part of the disk preferentially over others. A benchmark is the only way to really tell.

    3. Re:This is misleading. by Score+Whore · · Score: 1

      The time to wait for a sector to come under the head is known as latency. Seek time is the time to move the head.

      That said, I think you are confused even without the terminology mix up. It is true that there are more sectors in the tracks on the outer edge of the drive, however the platters are rigid and the speed is constant (I think WD recently released a drive with a variable spin rate, but most drives run at a constant 7200, 10k, or 15k RPM.) This means that whether a sector is on the innermost track, in the middle, or on the outermost track, it takes exactly the same amount of time to go around. The average latency is the same across the entire drive. It is always 1/2 the time it takes to go around once. A 7200 RPM drive is 4.1 ms, a 10K RPM drive is 3 ms, and a 15K RPM drive is 2 ms.

      Due to the higher linear sector density (not 2D bit/sector density since that is approximately constant (for very loose definitions of constant) across the entire drive) of the outer tracks, there are also fewer seeks required for any given amount of data.

      Interleave is only a method to hide the time it takes the drive electronics to finish with one sector and become ready for the next. From a performance perspective you can ignore it since the manufacturer will have laid out the sectors on the drive such that the next sequentially numbered sector is arriving at the head as soon as the electronics are ready to read it. For suitably quick electronics with moderately sized read caches there is no interleaving at all.

      The inner most tracks are slower or the same for every speed measurement of a drive: The latency is identical. The data rate is slower. And you get more head movement.

    4. Re:This is misleading. by Score+Whore · · Score: 1

      Due to the higher linear sector density

      I mean of course, the higher per track sector count.

  98. Re:To this whole chain of comments, I would like by AngelofDeath-02 · · Score: 1

    Maybe - but software raid has some advantages. 1) It's not tied to a particular piece of hardware to function. Meaning in the event of a catastrophic failure (motherboard dies, or raid controller card) Anything I have will work. 2) It is usable while rebuilding the raid, which the raid controller I have most certainly is not (And takes a great deal more time to build)

    --
    No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
  99. UPS by Technomancer · · Score: 1

    And I don't mean the shipping company.

  100. Benchmarks by yahwey · · Score: 1

    This is something I have been researching recently to store my photography, and I found that Toms Hardware has some good reviews and benchmarks using Intel's NAS performance toolkit. This review http://www.tomshardware.com/reviews/DS207-synology-nas,2081-5.html shows a comparison of the Synology DS-207 to similar lower priced units. From what I've seen, most of the inexpensive NAS's don't have very good throughput either due to low power cpu's, or slower bus speeds. I would like to have a nice appliance, but for the best price/performance it looks like I'll be building a box with FreeNAS or OpenFiler.

  101. Re:To this whole chain of comments, I would like by pyite · · Score: 3, Insightful

    ... to say that software RAID is almost invariably a poor solution. It is woefully slow compared to even a slow hardware RAID implementation.

    Spend a few bucks and get the right hardware. It is not expensive these days.

    This may have been true years ago, but it's not anymore. Modern CPUs can handle parity computations without a problem. As long as your controllers can support the throughput needed, there is no need for hardware RAID. After all, we have ZFS.

    Storage is undergoing a massive paradigm shift and folks like EMC are being caught with their pants down. Their spindle cost and price per GB is just too high.

    --

    "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

  102. Re:1Gb-T marketing gimmick, not speed rating by lpq · · Score: 1

    if it says gigabit ethernet, for me that usually means anywhere between 200-800Mbps
    And you would be one of those 'suckers' born every minute.

    I tried a 1Ge that allowed an internal 2-disk RAID-0, setup. The best it would do was about 12MB/s read, 7-8MB/s write.
    That's about the performance I'd expect on a 100Mb ethernet -- they just added Gb-E, as a marketing gimmick and because with current volumes -- GbE are probably getting chearper than 100bE.

    Do not assume 1Gb ethernet is >= 100Mb ethernet for throughput unless you have numbers to back it up.

    It's a very sad state of NAS units -- it's been only recently that they started offering 1Gb, but I have yet to see any that really support those speeds.

  103. Re:To this whole chain of comments, I would like by Anonymous Coward · · Score: 0

    On my contemporary $100 Intel desktop motherboard, with $170 CPU and three $90 sata drives, I get 165 MB/s via linux software raid5, for local access to large data files (such as I think the OP is mentioning for moving around test data). This is a 1.5 TB system costing me less than $600 USD, including the RAM and cheap case plus PSU. I use it for my DVR and it is absurdly over-provisioned.

    People need to understand that small file access is significantly affected by round-trip latency when using naive programs that do sequential access. The LAN-based access protocols will not pipeline file access and keep the LAN saturated if your program isn't issuing pipelined access to a stream of these small files...

    And those little SOHO NAS boxes are a joke, they cannot saturate the 100baseT speed, and are essentially false advertising when they add gigabit link speed and pretend it is better than the older model.

  104. Gig-E is slower than a single disk these days by metalhed77 · · Score: 1

    The days of network being faster than disk are over.

    Gig-E speed is about 30MB/s in the real world. This is with a crossover cable, machine to machine. I've tested and verified this over a number of platforms, including expensive server systems.

    Cheap terabyte single disks these days can do 80MB/s.

    The only reason to go with raid is for redundancy, or better handling of simultaneous connections.

    Read this for a more in depth analysis: http://www.codinghorror.com/blog/archives/000339.html

    --
    Photos.
    1. Re:Gig-E is slower than a single disk these days by pyite · · Score: 1

      Gig-E speed is about 30MB/s in the real world. This is with a crossover cable, machine to machine. I've tested and verified this over a number of platforms, including expensive server systems.

      Either your "real world" resides in a parallel universe where things aren't what they seem, or your test was flawed.

      I recently tested my MacBook Pro to my iMac over GigE through a switch (using iperf). With jumbo frames enabled, I got 986 Mbps. With jumbo frames disabled, I got about 870 Mbps. Not sure where your "real world" 30MB/s comes from.

      --

      "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

    2. Re:Gig-E is slower than a single disk these days by drsmithy · · Score: 1
      Rubbish:
      [ZUR (UTC+0100) csmith@dr01 /tmp]$
      scp dr02:/tmp/1000.testfile /tmp/
      1000.testfile
      100% 1000MB 47.6MB/s 00:21
      [ZUR (UTC+0100) csmith@dr01 /tmp]$

      That's through a switch, on a reasonably active network, without even the slightest attempt at tuning. I have little doubt I could push that up to 80+MB/sec without much effort (albeit probably not with SCP, which has other performance limitations) and I've gotten 150MB/sec with bonded NICs and iSCSI.

  105. Re:To this whole chain of comments, I would like by Kent+Recal · · Score: 3, Insightful

    software RAID ... is woefully slow compared to even a slow hardware RAID

    Wrong. Go do your homework.

  106. Samba is not slow by Anonymous Coward · · Score: 0

    I can't speak for your config. Mine is obviously different, cheap parts on my side. The only name brand anything are the Seagate disks.

    Sustained single disk speeds are around 12Mbps. Using RAID, you can increase that a little depending on the width of your array and the amount of disk cache RAM you have available.

    My SMB server is a Ubuntu 6.xx LTS desktop. It runs all sorts of other processes, but also has an external RAID5 array (4x320GB) connected via infiniband. The network is 1GB switched.

    I use Linux software RAID, not hardware RAID due to driver issues with my cheap Promise RAID card and the added flexibility it provides. Here's an excellent article for why using software RAID is a good idea http://linux.yyz.us/why-software-raid.html

    The server has only 2GB of RAM. Most of the RAM goes towards disk buffering - no tuning needed. I backup via rsync over ssh about 10GB of data every morning. 50Mbps throughput until the buffer gets filled, then it is limited by the disk performance on the source systems, usually around 12Mbps. Using network testing tools, I was able to see 550+Mbps throughput for pure network tests between Ubuntu clients and servers, so the network isn't an issue. WinXP client throughput was only slightly less over samba. I run WinXP in a Vbox VM under Vista, so that isn't really fair.

  107. Samba performance by ggendel · · Score: 1

    As for Samba performance... I have an OpenSolaris server which I use with a 2.5 TB disk stack in a Raid-5z configuration as a NAS for my SOHO setup. I set it up to use as Mac OS/X time-machine server for my Mac machines. With Samba, the initial backup took about 5 hours and consumed 1 of my two CPUs. I switched over to the native CIFS server in OpenSolaris and tried it from scratch again. This time it took under 2 hours and consumed less than 1% total CPU. I had similar experiences when I backed up my Windows machines. Sun still has the fastest NFS stack if that's the way you want to go as well. But, I typically use dirvish (rsync based) to back up my Unix/Linux machines nightly to the server. In 3 years of heavy use, I've never had a single hiccup.

  108. Make your own by jonny55555 · · Score: 1

    Yes and the MAKE webzine is pretty cool too!

    What makes me happy is a CentOS 5.2 server running a Promise ST EX8650 RAID-1 and RAID-5 solution (6 HDs on the SATA300 capable controller - oh, it does SAS too and is capable of addressing >200 HDs). You have to use ES.2 Seagate or similar drives otherwise you don't have their vendor support, but they run FAST and with my GiGe (Gigabit Ethernet) it runs at about 26 MBps. Currently my box is running Samba, NFS, and 3 VMWare systems because it has a nice AMD AM2 Dual Proc and a decent amount of RAM.

    This allows you to run server apps in the virtual systems without worrying about affecting your priceless data. Just turned up my Azureus virtual system recently and am able to turn off the main system in my room at night (save a little electricity now ^_^). With my setup, I actually added 2 10/100 cards and have them set to be bridged to the actual virtual machines (direct IP to the network).

    Specs
    AMD Athlon(tm) 64 X2 2.4 Ghz
    EPoX AM2+ Mobo
    1.5 GB DDR2 RAM
    1.1 TB Logical Drive 1
    1.4 TB Logical Drive 2
    2 GB Ethernet
    2 10/100 Ethernet

    Cost me around $800 (HDs and RAID card, the rest I already pretty well had)

    --
    Jonny5 'ko derf'
  109. Re:To this whole chain of comments, I would like by dfn_deux · · Score: 1

    I'm gonna call bullshit on this right now. General purpose processors have been dropping in price and going up in speed and core count much faster than hardware raid solutions. Take for example your average 12 port areca or 3ware or lsi raid card which will cost you 700 bucks, now instead of spending 700 on that card buy three 4 port lsi sata cards and spend the 300 bucks you save on a faster proc or one with a higher core density (4 vs 2 for example). You'll get not only faster or at least comparable performance, but you'll also get performance that scales up as you upgrade the rest of the system instead of being stuck with the perf that you get from your raid card. Another bonus is greater compatibility with different OS's as raid card drivers are traditionally notorious for having poor support in the long term and also have less tunability of the underlying logic. Finally you get better data portability, with a HW raid card you are stuck with the type of card you have used to create the array, with zfs or linux raid you are only limited to the OS you used; which sounds silly at first, but 3-4 years down the line when you are trying to find someone who stocks a defunct HW raid card to replace your burned out card you'll quickly appreciate. This is the exact same paradigm that makes things like SSL offload accelerators a poor value for a system which will be in use for more than a year or two, the more general use hardware will always outpace the performance of the purpose built stuff and by the nature of it being general use it will be far more flexible in the long run.

    --
    -*The above statement is printed entirely on recycled electrons*-
  110. Avoid Netgear by Macgruder · · Score: 1

    Avoid Netgear at all costs.... I've got two, once GigE, and one 100/E. They suck, not because of the throughput, but because of the super crappy Z-san service that you have to use to access them across the network.

    I've also got a couple of Western Digital My Book World II's, both GigE, and they work just peachy. I just finished tranferring 1Tb of data across my network, and it took about 36 hours. Bunches of 350 and 700 Mb files. They each take about 3 mins to transfer across the network.

    --
    I'm not crazy,I'm actively irresponsible.
  111. Qnap makes high speed NAS boxes by Anonymous Coward · · Score: 0

    I have been researching NAS appliances for some time and the top performers that I have found are made by Qnap http://www.qnap.com/product-index.asp

    Several sites have benchmarked some of their high end arrays around 60-70MB/s

  112. Sun 7000 series by Anonymous Coward · · Score: 0

    http://www.sun.com/storage/disk_systems/unified_storage/

  113. My numbers by Anonymous Coward · · Score: 0

    I have the buffalo you speak of and I only get 13MB up and 13.5MB down over gigE with 3 year old desktop with xp.

  114. Galactic constant by flyingfsck · · Score: 1

    The Galactic Constant for data transfer on a cheap X86 system seems to be 32MB/s. I've tested lots of cheap machines and they all turn out to max out at that speed, so good luck with your quest.

    --
    Excuse me, but please get off my Pennisetum Clandestinum, eh!
  115. Re:To this whole chain of comments, I would like by 1lus10n · · Score: 0

    EMC is actually doing quite well, and Netapp is kicking ass as well. The only shops that actually look at cost/GB as a measuring stick are small shops, or shops with very specific needs.

    Large corporations, government and high tech companies are usually more concerned with management costs, retention, migration and so forth.

    Not to mention getting into things like actual storage utilization - often thin provisioning and deduplication (see: Netapp) cuts your utilization so far down that you can purchase systems two or three models below what you were looking at before hand.

    I cannot imagine managing multiple sites with multiple TB of data inside of just a Linux or Solaris box. Talk about a nightmare.

    http://www.theregister.co.uk/2008/12/08/idc_q308_disk_storage_numbers/

    --
    "Two things are infinite: the universe and human stupidity; and I'm not sure about the the universe." --Albert Einstein
  116. IMO, NASes are way overrated by melted · · Score: 1

    I looked at the pricing and could not stomach it. A full blown Athlon 64 dual core box with 2GB of RAM costs less than a NAS box with no disks. And it can do a heck of a lot more. So that's what I ended up doing - bought an "energy efficient" version of Athlon 64, a mobo with integrated graphics, and stuffed it with SATA hard drives. RAID5, web server, file server, torrent, I even run scientific computations on it every now and then. I was running a VIA Epia box before it, but that one needed a SATA card to run RAID, since it only had two SATA ports. Best of all, I know that if the box goes titsup, I'll just replace the mobo and it will be able to run my RAID array again. The same can't be guaranteed with NAS boxes, since they use "proprietary" algorithms.

  117. Thecus N5200 by gora · · Score: 1

    I have Thecus N5200BR with 5 x 1TB WD Greenpower drives in a RAID 5 array. It uses a Linux based software raid but has a nice, relatively beefy CPU to keep speeds up. Supports Gigabit ethernet with jumbo frames. I believe you can couple the LAN and WAN ports to double your network bandwidth. You can also open it up and upgrade the RAM if you're adventurous enough. It supports the major file sharing services and I've had no problem with it on my home network (Vista 64 and 32, XP, OS X Leopard, OS X Tiger, Ubuntu). Works beautifully out the box, but you can install various "modules" as well to give you shell access, media servers, etc. All in all a fantastic high performance NAS.

  118. SmallNetBuilder by mr_death · · Score: 1
    --
    It's Linux, damnit! Pay no attention to renaming attempts by self-aggrandizing blowhards.
  119. Re:To this whole chain of comments, I would like by pyite · · Score: 3, Interesting

    The only shops that actually look at cost/GB as a measuring stick are small shops, or shops with very specific needs.

    Large corporations, government and high tech companies are usually more concerned with management costs, retention, migration and so forth.

    This is simply not true. There are plenty of commodity storage requirements that do not require Fibre Channel or even NetApp level NAS. On the other end of the spectrum, cost/GB might not be a huge factor, but the cost of getting necessary IOPS is certainly a factor.

    I work on Wall St. and we have multiple PB of storage. We have tons of EMC. However, things like the Sun X4500 and similar products from HP are changing the game. Couple that with being able to do 48 ports of line-rate 10GigE in a 1 RMU stackable, per priority pause coming into use, and Data Center Ethernet down the road and you have many reasons to seriously reconsider the scope of your fibre channel deployment.

    --

    "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

  120. iperf is not the real world, or even close to it by metalhed77 · · Score: 1

    I get 950 Mbps with iperf (w/o jumbo frames). Actual disk to disk, or even disk to memory transfer speeds are at 30MB/s. If you'd read the article I linked to you'd know that.

    Actual protocols are a LOT slower than just slamming as many bits as possible through a pipe.

    --
    Photos.
  121. Cheap Fast Reliable - Pick two :-) by Anonymous Coward · · Score: 0

    Go with something like Infiniband, or 10GB ether (should give you access to FCOE via ISCSI for instance. Fiber SCSI SAN the sucker, and there you go.
    Look up netapp.

    Enjoy.

  122. Business, don't go too cheap by guruevi · · Score: 1

    I work in an imaging department (fMRI). We're moving around usually gigabytes in datasets. If you need fast, simple, cheap storage for OS X we went with the Apple XRAID's which have now been replaced by the Promise machines. They're dirt-cheap for what they have to offer, they're plenty fast and provide all the safeguards (battery backup etc.) and tools you need. I am sometimes pumping 500 Mbps per machine over NFS and we are still on dual PowerPC G5's, the Intel's go even further.

    If you need specifics on our configuration let me know, just contact me, I made some tweaks to the TCP/IP settings to get this type of throughput.

    If you really want to go cheaper, consider the NORCO DS-1220. It's a 12-bay eSATA enclosure. It needs a bit tweaking to get it to work properly but it's plenty fast and dirt cheap (enclosure is under $800 + 10 disks) Only use the 10 disks on the expanders though if you're limited in a 1U server. You can use the 4-channel eSATA controller to drive 2 enclosures that way (20 drives)

    --
    Custom electronics and digital signage for your business: www.evcircuits.com
  123. TeraStation Pro (No) by troll8901 · · Score: 1

    My ex-company had a TeraStation Pro. Throughput was not good. It was also unable to do decent authentication. Flashed with the latest firmware 1.04 at that time.

    1. Re:TeraStation Pro (No) by Anonymous Coward · · Score: 0

      Second that. Throughput was sub 10MB/s through loads of different configurations. Turning off RAID, using them as 4 discreet drives and using NFS helped creep up to about 9.5.

      In the end I bought a cheap HP Server put the discs from the TeraStation in, installed FreeNAS, used software RAID 5 and now have a 1.5TB usable space NAS capable of 70MB/s.

  124. Re:To this whole chain of comments, I would like by Anonymous Coward · · Score: 1, Funny

    ... to say that software RAID is almost invariably a poor solution. It is woefully slow compared to even a slow hardware RAID implementation.

    Please cure your ignorance here.

    Or simply test some systems before you talk nonsense.

    Spend a few bucks and get the right hardware. It is not expensive these days.

    Are you selling raid controllers or are you one who actually believe their marketing bullshit?

  125. X Serve by Anonymous Coward · · Score: 1

    I know, I know, it's not a NAS, but I worked in the broadcast video world for a while and I started using them as NAS boxes since I needed to sustain 300 megabytes per second over bonded ethernet. It worked well.

    But on the other hand, an HP rack server with at least a quad core Xeon and OpenFiler is a GREAT/AWESOME option.

    Let me just point out what hasn't been said here so far. For sustained high speed data transfer, Xeon blows the doors off of Core 2 since the focus of those chips performance wise has been on bus speeds.

  126. Thecus N5200 Pro & N7700 by llin · · Score: 1

    To answer the poster's question, I have a Thecus N5200 Pro that performs pretty well - RAID0 should break 40MB/s - not amazing, but better than any other SOHO NAS's around, barring the new N7700, which looks like it'll hit Gig-E limits (finally a cheapish NAS that approaches 100MB/s for reads/writes). That said, you're gonna pay for the performance - the N5200 is about $700 and the N7700 is $1100 (enclosure's only). eAegis sells them w/ burned in drives as well - that's where I got my N5200. The hot-swap and automatic RAID rebuilding works as well and it has built in FTP, SMB, AFP, and NFS and is pretty good for a plug and play system.

    That being said, you're definitely paying a premium, and you could easily throw together a multi-terabyte system that would max out your GigE for about the same price as what you'd pay for the N5200 enclosure. My only big recommendation there is that you get a hotswap rack w/ that - makes things much more pleasant when replacing drives.

    One other thing to consider is power consumption. My N5200 Pro idles at 80W - if you built a low-power mini-ITX system you could probably get something pretty close to that, but a regular PC would probably be closer to 150-200W. Depending on your electric billing, you could be talking about a $100-200/yr difference there.

  127. ReadyNAS Duo by BlackCreek · · Score: 1
    I own a ReadyNAS Duo. While it has Gigabit network, my drives have a throughput of 30MBytes/s (which I find ridiculous since my much older and cheaper laptop disk is twice as fast). So regardless of the network throughput the internal slowness of (I guess) the motherboard will already hold you out.

    Some other comments:
    PROS: At least until the last OS iteration from Netgear they used a Debian based system. So you could just use apt to get new stuff installed. The Netgear forums are busy and provide useful answers to questions. The Duo specifically is small and silent, and was therefore approved by the girlfriend. Can be programmed to turn/off on specific time periods of the day, which is ideal for a home file server.

    CONS: The processor (sparc arch) is quite slow. The Debian dist they are based on is "sarge" (not maintained anymore), you can't upgrade the system (with another Debian system), and their own ReadyNas software is installed in a subsystem so changing its options through the command line remotely with ssh is not straightforward.

  128. Re: not anymore by phoenix321 · · Score: 2, Informative

    I have numbers to back it up: D-LINK DNS-323, 2x 500gb 5400rpm Samsung drives in Raid-1 configuration. I don't know the exact model, but I certainly selected these for low noise, low energy consumption and low heat output. So they're absolutely no high performers, but in regular, day-to-day operations, the Gigabit adapter manages a throughput at a steady 15 percent of 1000mbit push and pull from/to medium performance Windows workstations.

    This NAS unit is on the market for well over a year and it took several firmware revisions before other problems were worked out - but raw speed above 100mbit was never an issue. I don't have any real high performance client workstations, so I cannot say if these steady 150mbit throughput is limited by client or the NAS itself, but it certainly is enough to max out any and all WiFi links, which is enough for many applications except full disk backups, which take some hours in any case.

    I researched for a while before buying and got pretty much what other users described. I suggest you do the same so you can avoid the bad apples in the crowd of NAS units.

  129. Do it yourself by curious.corn · · Score: 1

    I second those that suggest building your own box. Just make sure you get all disk controllers on PCIe together with the NIC itself. Also forget about the integrated NIC and get a quality one, Broadcom or Intel with the latest drivers. Jumbo frames and TCP Offloading enabled will get your data across faster. Finally, if the switch supports it, put two or more NICs and enable bonding across them. As other posters suggested analyze your workload, the level of concurrent access and decide wheter to get many small spindles or less higher capacity.

    That was my -.02
       

    --
    Mi domando chi à il mandante di tutte le cazzate che faccio - Altan
  130. storevault by Exter-C · · Score: 0

    Hi There,

    I have been using a network appliance storevault in our environment that has similar requirements to what you seem to have. The device is a little more expensive when compared to the other netgear etc products you mention but for large sequential file transfers we will almost always get line-rate gigabit performance. It is great because it also gives you a gateway to experience the data ontap software from the netapp range. We paid just under £5000 for the particular setup we have but I have never looked back. Everyone in the organization is extremely happy with the products performance and reliability and at the end of the day isn't that what it is all about?

  131. something the wife will love by Anonymous Coward · · Score: 0

    just get one of these things and stick it next to the fridge in the kitchen.

    http://h18006.www1.hp.com/products/storageworks/xp24000/index.html

  132. PolyServe Matrix Server by Nagilum23 · · Score: 1

    It sounds like a perfect case for this except I'm not sure if there is a MacOS version. But you could still run it on the remaining Linux & Windows boxes and share it out from there.

  133. Re:UnRaid: when build-from-scratch isn't fast enou by jonaskoelker · · Score: 1

    costs $0 for a three disc license and $69(?) for a 6 disk license.

    I can't speak for anyone else, but you lost me there...

  134. Re:UnRaid: when build-from-scratch isn't fast enou by Anonymous Coward · · Score: 0

    unRAID is very, very picky with hardware. I've had so many issues both times I tried to setup unRAID that I've given up in it. I know people who can't say enough good things about it, but all I ever got were hard hangs and reiserfs errors.

  135. Synology DS408 by ImdatS · · Score: 2, Informative

    I used 2 LaCies for a while, but they both had a throughput of 10MB/s (the NAS with XP as OS) and 6MB/s (LaCie with Linux).

    Then I switched to Synology DS408. Mine has 4x Seagate 1.5TB HDs, RAID 5, so I have around 4TB of space.

    The network throughput maxes out at around 60MB/s(!). But this might be due to my not-so-good switch. It's all on a Gbps-Network.

    I used it only with Mac OS X (iMac, MBP, MBA, MB) with AFP. I haven't tested performance with SMB or NFS, but should be as fast as AFP (probably even faster).

    One thing, which really convinced me of Synology, was their support. Since the Seagate 1.5TB HDs have some problems (make sure you buy those with Firmware >=SD1A), I had a lot of issues at the beginning and thought that it's a problem with the NAS. I even thought I lost data. When I contacted Synology, they offered to log-on on to the NAS and try recovery, local check and everything - for free. And in the end, they found the problem with the Seagate HDs, proposed the solution and I am now even more happy then before.

    And no, I'm not working at Synology...

  136. super fast by Anonymous Coward · · Score: 0

    If you want the fastest possible speed, and if money is not the object, and you have a reasonably low amount of "streams" in and out, look at a Bright SAN. In the film world you can get multiple streams of uncompressed 2k dpx's playing off these without dropping frames, and they have gone a long way to solving the fragmetation issue that affects similar systems, particularly as the have written the FS.

  137. Software RAID is significantly faster by Kludge · · Score: 1

    I have always found software RAID to be faster than hardware. Apparently CPUs can XOR like crazy.
    I recently assembled a 4 disk software RAID-5 system for a friend. It read at 300 MBytes/sec, and no, it wasn't in cache. I thought that was pretty impressive.

    1. Re:Software RAID is significantly faster by Anonymous Coward · · Score: 0

      I can agree somewhat; we had twin xeon quad-core E5345 with Intel 5000 chipset with built-in sata, and s/w raid 5 was nearly as fast as raw SATA 3Gb/s. However, it did eat a whole core and a bit when running bonnie disk test.

      That said, we've found that the Dell perc 6/ir in a 2950ii with sata drives gives a very respectable performance, better than the above by up to a factor of three, better than a 3ware sata card by a factor of two, and similar ratio to the E200 in an HP DL-180.

      Posting a/c because I wouldn't want to upset HP who've been very nice to us :-)

    2. Re:Software RAID is significantly faster by SETIGuy · · Score: 1

      I have always found software RAID to be faster than hardware. Apparently CPUs can XOR like crazy. I recently assembled a 4 disk software RAID-5 system for a friend. It read at 300 MBytes/sec, and no, it wasn't in cache. I thought that was pretty impressive.

      You do realize that no XOR calculations are done when reading from a RAID-5 system (hardware or software) unless a drive has failed, right?

      But yes, in general, parity calculation for writing (or for reading from a degraded array) will be much faster on the host CPU than it would on the processor contained in a typical hardware RAID controller.

      What a hardware RAID controller does usually buy you (whether you use the onboard RAID or not) is battery backup of pending writes.

  138. Openfiler is good too by cciRRus · · Score: 2, Informative

    Instead of FreeNAS, I've tried . I managed to configure an iSCSI target with DRBD as the datastore for my VMware ESX 3.5 server.

    OpenFiler is neat and easy to use. Check it out too.

    --
    w00t
    1. Re:Openfiler is good too by jon3k · · Score: 1

      I'm using openfiler on an hp ml570 g3 with a pair of hp dl380 g4ps running esxi 3.5 in a lab and it's been phenomenal. Using some old scavenged fiber NIC's that don't even support jumbo frames I was able to get over 500mb/s to the iscsi target on the openfiler server. It's nice to be able to build your own makeshift san out of some spare hardware from the datacenter.

  139. General Info by Ozric · · Score: 1

    Thecus are good. You can get one that does i-scsi or has a GigE router.

    Qnap is very solid.

    Synology if you like features.

    Roll your own, a box with Direct attached storage is hard to beat.

  140. Alternatively, use iperf by cciRRus · · Score: 1

    Instead of using nc and pv to test the network throughput between a server and a client, try iperf.

    --
    w00t
    1. Re:Alternatively, use iperf by swillden · · Score: 1

      Check the thread. An AC already suggested that, and I responded.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
  141. Roll your own LinuxBox by JoeMerchant · · Score: 1

    I looked through the SoHo NAS market pretty thoroughly before settling on a QNAP TS-109 for my needs. It does NOT have the kind of bandwidth you are looking for, it is better than many of the other low-power boxes, but all of the low power boxes are CPU limited in terms of bandwidth.

    For the same money as a dedicated NAS box, you can get a low-end PC with Linux that will kick the NAS box to the curb on performance. The drawbacks of the Linux box are:

    1. Size
    2. Power Consumption (related: Noise and Heat)
    3. Configuration effort

    Size may not be a problem for you, and if you want 4 disk 7200RPM SATA II Raid 0 performance (or better), you're not going to be in a particularly small box anyway. Same goes for power consumption - 4 drives are starting to be quite a load in their own right. Which leaves: configuration effort - the NAS boxes are mostly plug and play, you might spend a couple of hours installing Debian (or similar) with Samba support, but after the initial setup, maintenance should be similar (backups, regular prayers to the spindle bearing and head crash deities).

    The place where a Linux box can get out of hand is its flexibility and configurability, you can always slap a new application server on there for some purpose or another - if you resist that temptation and keep it simple as a NAS, the Linux box should be every bit as reliable and trouble free as a NAS.

  142. If supported by NAS.. NetBeui.. by FirstOne · · Score: 2, Interesting

    If the NAS supports the non-routable NetBeui protocal.

    Install the optional "Netbeui" protocal stack located on the XP install disk. (same add-on will also work on Vista.)

    Don't forget to disable (uncheck) the "QOS Packet Scheduler", it will limit you to 20-25% of max link speed.

    Lastly, one must also disable the NetBIOS over TCP/IP, if it connects first you won't see any performance boost. (Option located in the TCP/IP Advanced/WINS dialog).

    The older/non-routable NetBeui protocal stack in the NT/W2K days was roughly 10x more CPU efficient per byte than NetBios over TCP/IP.

    In XP/Vista environments it's still 5x more CPU eff than NetBios over TCP/IP.

  143. SNAP Server by tarunbk · · Score: 1

    You can try snap server, it has enterprise level speed and reliability, flexibility... supports almost all the file sharing protocols and authentication protocols out there...

  144. Buffalo Linkstation NAS & HP Media Vault by montjoy0 · · Score: 1

    I currently own a Buffalo Linkstation NAS, it gets about 20 MB (megabytes) of traffic/sec on reads max, which isn't all that great but at least it's into GigE speeds. The web interface is decent, and since it supports samba and ftp it should work fine with OSX.
    I also have tried the HP Media Vault (their small linux NAS), but I found that when I put a second drive in it would over-heat and turn off. The user interface sucked too. I didn't get a chance to benchmark its speeds before I returned it.

    1. Re:Buffalo Linkstation NAS & HP Media Vault by montjoy0 · · Score: 1

      I should also mention that I use the Linkstation in a RAID-1. You might get better speeds with a RAID-0 or jbod, but then of course you don't get the data protection.

  145. Re:UnRaid: when build-from-scratch isn't fast enou by Overzeetop · · Score: 1

    What, at $69? How much time do you think it costs to set up a linux machine from scratch, including all the sharing necessary for a mixed environment? How much do you think a development person costs? In billable hours, most of these guys waste $69 worth of time on coffee breaks or taking a dump. I've set up two of these, and each one took about 30 minutes (exclusive of the hardware build, which was zero for the old Dell I had). The last time I installed ubuntu it took 30 minutes just to get to the startup boot.

    Free is not the end-all, be-all, especially in business. This would be $0 if he didn't need more than, say, 2TB of storage (3x1TB, one for parity), but $69 is not really a heavy cost. There are a lot of things I could do instead of paying for, but I don't, because even at my pedestrian $130/hr billing rate it's better for me to pay for a widget that fixes my problem than spend 4 hours troubleshooting it and getting it fixed myself.

    --
    Is it just my observation, or are there way too many stupid people in the world?
  146. Re:To this whole chain of comments, I would like by rthille · · Score: 1

    The only shops that actually look at cost/GB as a measuring stick are small shops, or shops with very specific needs.

    cough Google cough

    --
    Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
  147. Re:To this whole chain of comments, I would like by rthille · · Score: 1

    I don't think speed is the right reason to go with a hardware raid card like 3Ware. The real reason to go with a hardware raid card is for the battery backed cache. A UPS can help mitigate that issue, but that doesn't help with kernel panics. You still have to trust the software running on the raid card to do the right thing in the event of a crash of the OS, but being able to trust that the data you sent to memory on the card is on 'stable storage', run the drives with the write cache on (unsure how safe that is, even with the battery backed cache), and still be sure of your data is _huge_.

    --
    Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
  148. My FreeNAS performance by InvisiBill · · Score: 1

    I took an old Dell GX280 that we were getting rid of and turned it into a FreeNAS. The OS is installed on a 1GB MicroSD card in a tiny USB adapter and I stuck in a WD6400AAKS (WD 640GB SATA). Using the onboard NIC, I transferred a multiple-GB ISO between it and my PC (Athlon X2 6000+, 4GB, nForce 590, Seagate 500GB 7200.10). I sustained over 300Mb/s on the transfer. Also, my TrendNET switch between the two doesn't support Jumbo Frames, so this is with standard packet sizes. Not bad for two PCs with single SATA drives and onboard NICs, on a ~$120 NAS built in less than an hour (including hardware and software install).

  149. Nexenta/OpenSolaris by Anonymous Coward · · Score: 0

    If I had to build a NAS/SAN on the cheap for work it would be something based off of OpenSolaris/ZFS. The amount of features you'll get out of ZFS/Opensolaris for free can't be beat. Really worth a look. A few products that I would say to look at are: http://www.pogolinux.com/nexenta.php http://www.nexenta.com/corp/index.php?option=com_frontpage&Itemid=148 http://www.nexenta.org/os and of course: http://www.opensolaris.com/

  150. Is this true with Jumbo Frames? by Anonymous Coward · · Score: 0

    Is the theoritical limit on a Gig-E network still 119MBps with Jumbo-Frames turned on (MTU=9000)? Or does that theoritical limit go up to more (6x if liniar)? I ask as I just purchased a Jumbo-Frame capable Gig-E switch to use for a SAN: Ubuntu NFS box (amd 1.2GHz/1GB/4x600GB connected to a CPU-based PCI-4channel-Sata card.) Although it looks like I will also be PCI-limited in terms of throughput...

    1. Re:Is this true with Jumbo Frames? by swillden · · Score: 1

      Jumbo frames or not, the maximum throughput of a gigabit network is 10^9 bits per second. Assuming zero overhead, divide that by 8 to get 125 MBps, or, in the base-2 units we normally use for file sizes, 119.2 MiBps.

      Jumbo frames just allow those billion bits to be divided into fewer, larger, packets. That reduces header overhead and lets you use a few more of the billion bits for data.

      --
      Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
  151. 12-20 by Anonymous Coward · · Score: 0

    Typically you can expect between 10-20MB/s from a Marvell Orion powered consumer NAS, which is barely quicker than fast ethernet.

    Most consumer NAS these days use this SOC, with the exception of Thecus who cant seem to decide what vendor to standardise upon.

    If your storage must be NAS, build your own for better performance.

    Even when you go up to 4 bay NAS orion/powerpc based NAS they are still slow, and costing about £400-£500 which is a total rip off for the performance they give - for that you can build a very nice little storage box possible even with real hardware RAID.

  152. Do you REALLY need a NAS? by ugmoe2000 · · Score: 1

    It sounds to me like you need to better define the criteria which require a NAS. Whenever looking into new equipment it's always helpful to define the ideal situation in which your new equipment would function... for me based on what you mentioned I would say the ideal would be:

    -0 wait time for transfer
    -Portability of data

    At no point did you mention other people needing the data while you're running the tests. Based on those criteria I would shy away from a NAS. Unless you're willing to shell out a grand for a Linux solution, a 1 TB external drive with an eSata connection will be ~$900 cheaper and possibly perform better to those criteria above. When you're not doing the testing you can simply plug it into an idle machine for people access the data over the network and to perform backups.

    I'm a big fan of home-brewed Linux Raid NAS solutions, I setup and maintain a few myself for several different organizations and I regularly transfer 100+ gig dumps of files... however I would avoid doing all of this if I could. Keep it simple... it's much easier to just carry the data with you from machine to machine if your files are exceedingly large... after all HD's are cheap now-a-days.

    On another note if you do insist on a NAS, one thing I've noticed is that Filesizes can easily affect transfer rates. The smaller the average filesize, the larger the individual overhead is in comparison to the file. In other words if you're transferring 100meg files and the overhead is a theoretical 1kb then the majority of time is spent moving the file but if you're transferring 1kb files and the overhead is still 1kb then you spend just as much time on the overhead as on the files themselves.

    One trick I've learned to speed large transfers of small files along is to turn your files into a storage-level-compression tarball on the fly with a blowfish cipher and pipe the output over ssh to the other machine, where it is disassembled on the fly as well. This means that you're only transferring one file across the pipe and there is less overhead to the total transfer. This trick keeps the transfer rate on my machines steady at around 10-20mb/s as opposed to 6-7mb/s and those numbers are on a standard configuration with no special hd's or raid implemented. I can only expect the numbers to be better with raid. Just things to think about.

  153. An empirical answer to your question.... by Anonymous Coward · · Score: 0

    1. Go to www.google.com
    2. Type in the search box "nas throughput chart"
    3. Hit "I'm Feeling Lucky"

    There, question answered, you have to really master google search to find esoteric tech info such as this.

    (Notice that the charts displayed are on 100Mbsp network, you'll want to change the drop-down to 1000Mbps)

  154. CoRaid by Em+Ellel · · Score: 1

    First of all for the purposes mentioned in the original post, a USB HDD is by far the fastest, cheapest and most reliable solution. Transfer rates of 500MB in as long as it takes to walk from one desk to another.

    But if you are looking for a network based solution (and have $10k), I am surprised no one has mentioned CoRaid. CoRaid is a AoE SAN hardware company but they offer NAS solutions too (basically a 1u linux box connected to their SAN) - for about $8-9k ($6.6k+15 1TB drives) you can have a 15TB (unraided capacity) NAS with advertised throughput of 100MB/60MB (read/write).

    -Em

    --
    RelevantElephants: A Somatic WebComic...
  155. Expensive but by C_Kode · · Score: 1

    Use a SAN and you just zone the storage to the server that needs it when they need it. No transfering of GBs of data required. :)

  156. Aim Higher. by dghcasp · · Score: 1

    If performance is your goal, don't look at the "buy-it-at-Frys" level of NAS, or at "roll-your-own."

    If you build your own, you'll end up bottle-necked by the performance of the particular OS you use, plus SAMBA or NFS (depending on your needs.) Plus, there's the time factor in putting it together, tuning it, and maintaining it. Granted, this isn't a lot if you're already a tech-head, but your time isn't free.

    If you buy one of the consumer-level NAS boxes, what you're getting is the equivalent of building your own, without the ability to tune it, since most are based on the same open-source software you would use yourself. Pretty much every NAS device I've ever seen has the same cyclical bursty transfer profile as a build-it-yourself.

    If you want better performance and buzzwords, every major PC vendor now has a SAN solution. You get the benefit of a team of people whose job it is to maximize disk performance, and a nice management system. However, your system head still has the same problems as before - using a general purpose OS and/or open-source software.

    If you want pure performance, look at Network Appliance. They've been in the game for a long time, and their hardware/software combination allows them to control/tune the whole environment. To a first approximation, all the cool things in ZFS were done ten years ago by NetApp. You get the benefit of a whole company whose job it is to maximize disk and network performance. You can look at a performance review from earlier this year showing about 30k SPC-1 IOPS.

    Personal anecdotes:

    1. Several years ago, we did a benchmark for ClearCase between a Sun hardware head with a (a) directly connected, fibrechannel SCSI RAID array, and (b) a 100G ethernet connection to a NetApp. The performance of the NetApp was about 20% higher than directly connected disk.
    2. NetApp service is incredible. We came in one morning, and there was an email from Network Appliance that basically said "Hello; your NetApp notified us that one of its disks has failed. We have shipped a replacement. Here's your UPS tracking number."
    3. The above also holds for software. There's nothing like "Hello; your NetApp had a software failure. From analyzing the crash dump it sent us, we recommend you install patch xxyyzzz."

    Note: My only relation with NetApp is being a very satisfied customer.

  157. Info on Netgear ReadyNAS+ by cheddarlump · · Score: 1

    For what it's worth, I'm using a Netgear readynas+ here at work with 4 500GB drives in it (RAID5). I get usable throughput, but it's no PC. Reads are about 20-25MB/s on GigE without jumbo frames, writes are at around 12-15MB/s. One thing I really like about the box is the ability to SSH to it and manipulate files locally, I use it for holding security camera storage, and being able to mv 1000's of small files around is nice. Cost is high for what it is though, if I remember correctly, it's about $900 for the device with 2 drives in it. User interface is decent.

  158. thumbs up for Sans Digit by magnosis · · Score: 1

    As posted previously, we've tested a Thecus NAS and it was hell. We've since replaced it with a SansDigital MN4L+B, with 4x 750GB Seagate ES.2, and it works like a charm. It's the fastest NAS you'll find below 1000$ (beside building your own). It has a Celeron 1.5GHz processor, extra slots for expanding DDR2 RAM up to 2GB, and it's made of metal, some some cheap plastic junk like Thecus') It's also one of the very few NAS that support NFS and iSCSI. Yes, we were surprised indeed to find out that a LOT of NAS out there don't have NFS support. We have it configured as RAID 10, and it's as fast as you can get on the GbE link. It has also rebuilt our 1.3TB array in a matter of hours after a drive was replaced. So far, we're very happy with it. Note that I'm not comparing with the more expensive solutions (Synology, Intel, Dell...) although I doubt they bring much more to the table other than extra storage space ans SCSI support.

  159. Re:To this whole chain of comments, I would like by jp10558 · · Score: 1

    Doesn't stuff like ZFS pretty much make that a moot point?

    --
    Opera, Proxomitron-Grypen,GPG 0x0A1C6EE3
  160. Re:To this whole chain of comments, I would like by jon3k · · Score: 1

    Who's doing 48 ports of 10GbE? That's pretty damn impressive.

  161. unRAID by jon3k · · Score: 1

    Lime-Technology makes a really interesting piece of software. It's a bootable NAS using their raid implementation called unRAID, which is similar to RAID4 except that each disk has an independent filesystem. So, you can pull a single drive out and it's readable. Also, you can mix and match any drive sizes you like, the only condition being the largest drive is used for parity.

    Also the basic version is free

    http://lime-technology.com/

  162. Check out Intel SS4200-E NAS.. by Glasswire · · Score: 1

    4 drives and eSATA for expansion. Details here
    This uses a 1.6GHz Celeron, which, while low end for a general cpu, is pretty fast compared to most cheap NAS units and will come closer to utilizing the gig port than most.

  163. Re:To this whole chain of comments, I would like by rthille · · Score: 1

    I don't think so, if my mail server issues a write to ZFS, and it returns that the data is on 'stable storage' (fsync symantics), then issues the "250 message received" to the client, then immediately crashes; either 1 the message is lost if the data wasn't really on 'stable storage', or 2 ZFS had to wait for the write to complete to the platters or the battery-backed cache on the raid card. I'm guessing just having to get to the battery-backed cache on the raid card would speed things up quite a bit...

    --
    Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
  164. Re:Synology boxes - recommended by AliasMarlowe · · Score: 1

    I'll also recommend the Synology boxes, based on several years with them.

    They are typical SOHO NAS solutions, so they don't compete in performance with custom roll-your-own solutions or with a midrange (expensive) NAS. However, they do come with built-in RAID support. Most of them have built-in webserver, photo station, media server, and print server functions.

    The down side is that every Synology NAS assumes all of its clients are either Windows or Mac boxes. The supplied backup and p2p download functions only work with those clients. However, Synology provides free add-ons to upgrade their boxes to use NFS instead of SMB and add telnet and ssh support. We've had no problems using them in a Linux-only home.

    The administration interface works fine in Opera, Epiphany, and Firefox. It is supposed to work with Internet Explorer, but I cannot personnally confirm that.

    --
    Those who can make you believe absurdities can make you commit atrocities. - Voltaire
  165. Addonics - homebuilt by Anonymous Coward · · Score: 0

    My homebuilt NAS (not really NAS, just a desktop with an external RAID5 array) just finished receiving backup mirrors from another machine in the house (a Xen box with 6 VMs running). These transfers are over rsync/ssh with no pre-existing file to help with speed up. Only the seagate drives are branded, everything else is cheap. Software RAID5 on the target with about 1GB of disk buffer RAM, single disk on the source. GigE network with either built-in NICs or $9 cards. Ubuntu and JFS on both sides.

    crm46-20081217.tgz 478555089 100% 49.85MB/s 0:00:09 (xfer#1, to-check=45/54)
    dms44-20081217.tgz 3043217581 100% 16.49MB/s 0:02:55 (xfer#2, to-check=36/54)
    mon45-20081217.tgz 464314500 100% 43.80MB/s 0:00:10 (xfer#3, to-check=27/54)
    pki42-20081217.tgz 369984893 100% 40.45MB/s 0:00:08 (xfer#4, to-check=18/54)
    xen41-20081217.tgz 1546169689 100% 14.87MB/s 0:01:39 (xfer#5, to-check=9/54)
    zcs43-20081217.tgz 2496573639 100% 15.32MB/s 0:02:35 (xfer#6, to-check=0/54)
    total: matches=0 hash_hits=0 false_alarms=0 data=8398815391

    49MB/s = 392 Mbps
    15MB/s = 120 Mbps

    Not bad for cheap solution that's been working 3 years now. Cost was about $550 total. Check out the Addonics - http://addonics.com/products/raid_system/mst4.asp

  166. USB by zerkon · · Score: 1

    I know this doesn't exactly answer your question... but unless your machines are widely geographically separated, why not cut out the middleman and just get a hard drive and a usb2 enclosure? You know exactly what you are getting (instead of all the debate about "real world" performance). Sneakernet ftw

  167. Re:To this whole chain of comments, I would like by speculatrix · · Score: 1
  168. Re:To this whole chain of comments, I would like by WuphonsReach · · Score: 1

    software RAID ... is woefully slow compared to even a slow hardware RAID

    Wrong. Go do your homework.


    Well, maybe... what I've found over the years is that RAID-0, RAID-1, RAID-10 all perform extremely well as Software RAID without taxing the CPU.

    But RAID-5 and RAID-6 do tend to become CPU-bottlenecked given enough spindles.

    So if you're going with RAID-5 or RAID-6, makes sure to purchase a CPU with a high single-core performance (Linux Software RAID is currently not multi-threaded enough to split the calculations across multiple cores yet for a single array). The last RAID-6 box that I built, we made the mistake of going with a 1.8GHz CPU instead of a 2.4GHz or 2.6GHz CPU, which cut our performance by about 1/4 to 1/3 of what it could have been.

    The spindles are definitely not at 100% utilization, but you can see the software RAID process pegging one of the CPUs at 100%. (Fairly modern motherboard, PCIe-based, pair of Opteron dual-core, 8GB RAM, 15 SATA drives.)

    --
    Wolde you bothe eate your cake, and have your cake?
  169. Re:To this whole chain of comments, I would like by Kent+Recal · · Score: 1

    Well, that sounds better.
    Yes, RAID5 is taxing the CPU but then again you don't normally use RAID5 in a box that is supposed to host I/O intensive apps.

    About a year ago we benchmarked a few hardware RAID cards (3ware, LSI, adaptec) versus linux mdraid and found that only the very high end cards ($1000 range) could actually saturate our 16 SCA disks.
    mdraid had no problems maxing out the disks (on Dual Xeons iirc), admittedly under significant CPU load.

  170. Re:To this whole chain of comments, I would like by AngelofDeath-02 · · Score: 1

    It's supposed to tolerate these faults for modifying data because it writes additional or replacement information somewhere else and then updates the index. So - you still might lose the data, but you won't have your index pointing at something corrupt due to a partial write.

    --
    No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
  171. Re:To this whole chain of comments, I would like by HBI · · Score: 1

    Firm rule: those who malign hardware RAID aren't using the right controllers. The three you list are nowhere near the top of the heap. Try a HP Smart controller for some hardware RAID pleasure.

    Then again you were defending the Sun x4100 servers last time I saw you, which also has a POS RAID controller. And crappy disk bandwidth in general.

    --
    HBI's Law: Frequency of calling others Nazis is directly correlated with the likelihood of the accuser being Communist.
  172. Re:To this whole chain of comments, I would like by Kent+Recal · · Score: 1

    Well, I wouldn't even malign hardware raid for performance alone. Hardware raid is a bad idea for much better reasons, such as reliance on single vendors, non-portable on-disk formats, unknown failure modes etc. And RAID5, soft or hard, is a horrible idea with today's disk sizes anyways.

    And what's wrong with the PCIe performance in the xfires? We haven't seen any problems so far.
    If you buy their POS controller instead of your "HP pleasure" then who's to blame?

  173. Thecus NAS is best - why ? by Anonymous Coward · · Score: 0

    100% GPL - download the source

    N5200B Pro - easily over 70MB/s up to 7.5TB

    N7700/N8800 - easily over 100MB/s up to 10.5TB/12TB & ZFS & XFS

    Supports everything: Windows, OSX, Linux, BSD, Unix

    Supports everything: CIFS/SMB, AFP 3, NFSv3, FTP, HTTP, HTTPs

  174. Re:To this whole chain of comments, I would like by pyite · · Score: 1

    Who's doing 48 ports of 10GbE? That's pretty damn impressive.

    Arista Networks. Formally known as Arastra.

    --

    "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

  175. Re:To this whole chain of comments, I would like by pyite · · Score: 1

    Arista Networks. Formally known as Arastra.

    Self-correction: That should be formerly.

    --

    "Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman

  176. That was my point. by Jane+Q.+Public · · Score: 1

    Certainly you can do decent Raid via software, but (normall) why would you want to? That is to say, why load down the CPU when you can offload the task to dedicated hardware?

    If it is a box dedicated to JUST network storage, I could understand... and I suppose that might be what he had in mind. But otherwise you are misusing your resources.

    1. Re:That was my point. by WuphonsReach · · Score: 1

      A lot of it is not having to deal with the vagaries of finding the drivers to setup hardware RAID on a Linux box. Or deal with strange management interfaces, whereas if we simply use the RAID controller in JBOD mode, we can simply admin the box using Linux Software RAID, no matter what controller we use. And we can move those disks to another box, using a different RAID controller, and still have it work.

      Software RAID also allows you to slice up the spindles into strange and varied shapes. Such as mixing RAID6, RAID10, RAID1 across the same set of spindles, depending on your access patterns or sensitivity to data loss. I might assign the first half of 3 drives to RAID1, the first half of the next 7 drives to a RAID10 array with hot spare, and put a RAID6 across the first half of the last 5 drives. Then across the back half of all the drives, I can put a large, seldom-used RAID6 array for bulk storage.

      A lot of failure modes nowadays seem to be RAID slices deciding to desync with the rest of the array. Or bitrot which gets discovered during a weekly resync of the array. If a slice is an entire disk, you could be dealing with a large amount of resync work. But if you divide each disk into 4 slices and setup 4 arrays, a desync event might only affect one out of the four arrays. Giving you faster rebuild times.

      --
      Wolde you bothe eate your cake, and have your cake?
    2. Re:That was my point. by Kent+Recal · · Score: 1

      Certainly you can do decent Raid via software, but (normall) why would you want to? That is to say, why load down the CPU when you can offload the task to dedicated hardware?

      Well, first and foremost because you get a documented, open on-disk format which offers better chances for recovery when your raidset goes belly up in a really bad way (multi-disk failure, controller writing crap, silent bit rot, etc.). With a softraid you can just pop the remaining disks into an entirely new machine and start working on it with standard tools. With a hardware card: Good luck asking 3ware, LSI or HP for data recovery services after one of their controllers busted up.

      In practice, even porting a perfectly sync'ed RAID-set from one machine to another is a bit of a gamble as compatibility even between minor BIOS-revisions varies. And better forget about moving a degraded or otherwise damaged RAID anywhere...

      Furthermore softraid trumps all but the very high end cards in RAID-10 performance. A modern multi-GHZ CPU won't break a sweat over a few xor's - not so the flimsy dedicated CPUs on many raid cards.
      And since RAID10 is the only raid-level that's still interesting nowadays I don't see many use cases for hardware-raid anymore.

    3. Re:That was my point. by WuphonsReach · · Score: 1

      One other reason that I like software RAID.

      Let's say that we start with (15) 500GB disks. Next year I start dropping in 1TB disks, one-by-one. At first, we only use 500GB on the 1TB disks. But once we drop the 15th disk in, we can create additional partitions on the drives, create new Software RAID arrays, and then extend LVM volume groups across the new arrays.

      Which lets us double our data density, without having to backup lots of data, tear the server down, put new disks in, build an all new array, then restore all that data. (Although I would hope that modern hardware RAID card management would allow you to grow the array by changing out disks.)

      We might even upgrade the disks in the array a second time once 2TB drives get below $100 each. Past that point, and I'd expect the server to be getting long in the tooth with a need to be replaced.

      --
      Wolde you bothe eate your cake, and have your cake?
    4. Re:That was my point. by drsmithy · · Score: 1

      Certainly you can do decent Raid via software, but (normall) why would you want to? That is to say, why load down the CPU when you can offload the task to dedicated hardware?

      Because the general purpose CPU will do it faster and cheaper.

      To say nothing of the overall performance, reliability and flexibility advantages of software RAID.

      There are times and places to use hardware RAID. Booting from (for simplicity) and when the machine is bus-limited are the two major ones, but with modern PCIe machines the latter is rapidly becoming irrelevant unless you have massive numbers of spindles connected to a single box.

  177. Re:To this whole chain of comments, I would like by Jane+Q.+Public · · Score: 1

    Again that's fine, IF you want to dedicate your general-purpose processors for such tasks. But that is like shooting an ant with a cannon.

  178. Not confused by Jane+Q.+Public · · Score: 1

    I just did not want to get into the really gory details.

    But in fact if a drive is physically interleaved (which as you point out is invisible to the outside), the latency is NOT identical. A drive with interleaved sectors takes more than one physical revolution to read all the sectors on the "virtual" track; how many revolutions depends on the particular interleaving scheme.

    As you point out, the latency is related to how long it takes a disk to do one revolution. But on an interleaved track, a "virtual" revolution may be several physical revolutions.

    This might not be a common scheme anymore, but again, a benchmark is the only way to really tell.

    1. Re:Not confused by Anonymous Coward · · Score: 0

      "I just did not want to get into the really gory details."

      I noticed.

    2. Re:Not confused by Score+Whore · · Score: 1

      The latency is identical. The definition of latency is how long it takes a particular sector to become available. An interleave of 1:1, 1:3, 1:9, it doesn't matter. The maximum number of physical revolutions it takes to get to any particular sector is slightly less than one (one single bit's width less than one revolution actually.) When it comes to locating a sector your idea of a "virtual" revolution is irrelevant.

      Finally, you can't by a drive manufactured in the last ten years that is interleaved. They just don't exist. From IDE onward the electronics have been fast enough that there is no need.

  179. Re:To this whole chain of comments, I would like by jon3k · · Score: 1

    Good lord thats amazing. I had no idea that kind of port density existed.

  180. Re:To this whole chain of comments, I would like by rthille · · Score: 1

    Right, the filesystem wouldn't be corrupted, but data would be lost. Lots of people find that unacceptable :-)

    --
    Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
  181. Re:To this whole chain of comments, I would like by dfn_deux · · Score: 1

    Did you read what I wrote? IF you save money on the disk subsystem and use that money to buy more processor THEN you can use that additional processor to do a better job of managing your disk subsystem than the dedicated hardware could. So if you play it like I've outlined then you end up with faster disk when you need it and you also end up with more processor for other tasks when disk isn't taxing it.

    --
    -*The above statement is printed entirely on recycled electrons*-
  182. I stand corrected. by Jane+Q.+Public · · Score: 1

    You are right: it doesn't matter... today. It used to. I can cite one situation where the latency most definitely was not the same, though: when the interleave was incorrect. It could take many revolutions to find the correct sector.

    But that was due to improper low-level formatting and doesn't really count. :o) Although I did see some instances of it.

  183. Re:1Gb-T marketing gimmick, not speed rating by mathew7 · · Score: 1

    Do not assume 1Gb ethernet is >= 100Mb ethernet for throughput unless you have numbers to back it up.

    The problem with ethernet at 1Gbyte speeds was the limit of 1500-byte ethernet frame. So they added 9KB jumbo frames, but these are not supported by early devices (switches or cards). So if you do not use jumbo frames, you can easily bring down your cheap switch.

    I can tell you that I am using one router (PC with linux) with 1 laptop 7200 HDD (for torrents) and my gaming rig with no RAID in any of them. I can transfer 20-25MB/s because that is the limit of the 2.5" 7200rpm HDD. So my switch (TP-link 8-port all-gigabit) throught which all my network is connected has the bandwidth. But this is with big files (100+MB).

    For small files the file-system overhead and incomplete ethernet/IP packets have a big penalty to network speeds AND HDD reads/writes. Also, the busier the network is, the worst the performance is. Keep in mind that the higher latency you have, the lower speed you're gonna get. Especially with windows file sharing (SMB). While experimenting with a local cascading 100Mbps network, I saw that SMB was affected if you communicated through 2 switches (10MB/s) vs. 4 switches (7-8MB/s). The Server (test source) was the same but the client did not affect the result, because for FTP I got same speeds (10MB/s at 2 or 4 switches). There was one test file around 600MB in size.

    I tried a 1Ge that allowed an internal 2-disk RAID-0, setup. The best it would do was about 12MB/s read, 7-8MB/s write.

    From what I understand only one "device" had RAID setup. So maybe the other end (which was not RAID) could not handle more speed. Remember that when you copy, you have the speed of the weakest link. You would not expect to copy at RAID speed from a CD/DVD.

  184. Bad idea ! by DrYak · · Score: 1

    I wounder if you could use the GPU on that board to off load checksums and maybe encryption?

    This is definitely a *BAD* idea.
    Although AMD is active a lot inside the OpenCL development - which mean you could easily have a nice C99-like environment to develop your GPGPU-accelerated RAID system - there are 4 main reasons NOT to do it :

    1. The first reason is bandwidth :
      That mean you have to use twice the bandwith :
      once to load the data from harddisk to main RAM, and then a second time to upload the data from main RAM to Video RAM.
      Although, probably, with some clever programming trick you could peer-2-peer connect the GPU to the disk controller over PCIe, but this goes beyond the basic "throw a couple of code lines in C". And I don't know how the onboard GPU is internally connected to the rest of the bus in a 780G (does it have 16 dedicated PCIe lane - in which case it should be able to hold the load ? Or does it have a smaller connect which will bring an additional bottle neck to your setup ?)
    2. GPUs are designed with speed over anything else in mind. Including over accuracy. See for example the accuracy limits explained in the OpenCL documentation. It doesn't pose any form of problem for graphics (if one single pixel is the wrong colour during one single frame once in a while nobody will notice any way). It happens seldom enough not to be too much of a problem in scientific parallel calculations (and methods are developed to make up for the lack of precision anyway).
      But for something like RAID, whose *purpose* is to ensure reliability, and whose main enemy is bitrot, adding yet another layer that can break once in a while isn't the best practice.
      Although you can offload checksum *verifications* to the GPU and use the failures not to mark failed blocks, but blocks which will have to be double-checked on the CPU.
    3. GPU are specially designed for massive FLOATING-POINT computations. They're not optimized for other kind of computations (hence the PR/Marketing duel between Intel and nVidia about Larabee-accelerated Ray-tracing). For example, RV670 (Radeon 38x0) advertises "320 processors" - these are 80 blocs which each have 1 general purpose unit and 4 multiply-and-add units. That's indeed up to 640 floating point operations in parallel, but for RAID you needs "XOR"s and this chip will probably only perform 80 of them in parallel.
      Then, on the other hand, if your purpose isn't to speed up the process, but simply unload the CPU, you probably will manage to have more CPU time available while the GPU takes care of it.
    4. Last : it is best to compute checksums in the final stage where the data will be subsequently used : inside the CPU + RAM (usually: parity checked RAM in critical systems).
      With software RAID you have the absolute certainty that the copy you have in RAM is correct (and that copy will be kept subsequently correct thank to ECC RAM).
      With a hardware RAID, you have the risk that errors appear between the place were intergrity was checked (the hardware controller) and the place where it will be used (CPU + RAM) : and not all bus have error checking. PCI and PCI-X do not. The full PCIe has 32bit CRC, but not the "Graphics PCIe" used by most consumer material.
      Then again it depends on the degree of reliability you expect from your RAID configuration

    So, for a home server whose purpose is mainly to host the collection of photos and videos, a GPGPU RAID implementation might be a fun project, but I would seriously avoid this on a system requiring high reliability.

    --
    "Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
  185. Re:1Gb-T marketing gimmick, not speed rating by lpq · · Score: 1

    On non NAS devices (like from a linux server), I get 200-400Mb/s Xfer over 1G (w/non-raid disks). I don't see that on NAS devices.
    Max speed I was able to get with standard packets (1500 bytes) was about 700Mbs -- that's expectable/acceptable. I would "love" to use jumbo frames, but Jumbo Frames usage is flawed as currently implemented (IMO).

    Currently, all equipment must run at same frame size -- you can't upgrade "some cards", and switch over to larger packetsizes, slowly as you need to replace things...its all 1 size or nothing. The bad thing is that when larger packet devices would try to talk to smaller devices, they would work find for small packet xfers -- like 'ssh' -- but just "drop" larger packets on a TCP connection --
    no error sent back to sending device to break packet down into smaller chunks -- packet is just dropped.

    There was no negotiation (or no "correct negotiation") of max packet size -- if a host was set to allow use of large packets, it would use them talking to every client -- was MAJORLY bummed when I figured this out -- there was no way to tell the host to use different packet sizes based on some agreed max-MTU size -- all or nothing.

    I might have been able to use multiple virtual IP interfaces and cut up the network into different virtual, overlapping IP spaces based on their MAX MTU size, but that was way too much headache to enable Jumbo frames. Having to move clients around on a virtual subnet based on what card they were running and what Max-MTU they were capable of... There needs to be some auto-negotiation as is done for link-speed&duplex.

    Under test conditions, though, Jumbo packets only gave me about a max of 15% more throughput. In practice, it wasn't noticeable, but admittedly, could have been if my server had been using RAID or if more of the clients were linux (instead of Windows -- Windows stack was notably slower).

  186. Re:1Gb-T marketing gimmick, not speed rating by mathew7 · · Score: 1

    I don't know the details of the jumbo frames (like negotiations, if any), but I do know that at least the communicating points (hosts and switch) need to "know" it. But I see it as mainly a "switch helper". Think of an 8-point switch with 8 hosts and all of them transferring at maximum speed. So if you can send 1 big packet instead of 6, you can help the switch very much. Although the ehternet checksum is done on the same size (wether it's 6 1.5K packets or 1 9K), the overhead and switching are done 6-times less.
    So the jumbo frames will not give you much higher speeds (mainly the overhead 18 bytes/packet of the 5 packets that are sent if jumbo frames are disabled) unless your network is congested.
    But like you said: compatibility within the network is questionable.
    In my previous post I said that there was a problem (although I did not say it was about usage and not bandwidth), but not that the jumbo frames will give you always a speed boost.

  187. Re:To this whole chain of comments, I would like by AngelofDeath-02 · · Score: 1

    I guess it has more to do with what you're storing. Only recent data would be lost, so you wouldn't lose the entire document - just the latest revision. Really - that's an issue for any pending writes during a shutdown.

    --
    No, I am not an English major. My posts are subject to typos and incorrect grammar. Do not expect perfection.
  188. S550 is your friend by captjohnmiller · · Score: 1

    S550 from NetApp should serve your needs if it fits in your budget.

  189. Re:To this whole chain of comments, I would like by drsmithy · · Score: 1

    But RAID-5 and RAID-6 do tend to become CPU-bottlenecked given enough spindles.

    Even a 10+ year old 300Mhz Pentium 2 class CPU has a RAID5/6 checksumming speed well into the hundreds of MB/sec. It's not going to be a limiting factor.

    So if you're going with RAID-5 or RAID-6, makes sure to purchase a CPU with a high single-core performance (Linux Software RAID is currently not multi-threaded enough to split the calculations across multiple cores yet for a single array). The last RAID-6 box that I built, we made the mistake of going with a 1.8GHz CPU instead of a 2.4GHz or 2.6GHz CPU, which cut our performance by about 1/4 to 1/3 of what it could have been.

    The checksumming speed of the CPU was almost certainly not the problem (unless you were seriously loading down the machine with other processing). You don't mention what kind of CPU, but even a 1.8Ghz Pentium 4 will be able to compute RAID5/6 checksums in excess of a couple of gigabytes/sec. Most likely you were bus-limited, or had crappy disk controllers that couldn't shift data fast enough or were generating storms of interrupts.

    The "high CPU usage" of parity-based RAID is a myth that needs to be laid to rest (although it is good for identifying which people really do know what they're talking about when it comes to RAID - I often use it as an interview question).