Best Solutions For Massive Home Hard Drive Storage?

i_ate_god writes "I download a lot of 720/1080p videos, and I also produce a lot of raw uncompressed video. I have run out of slots to put in hard drives across two computers. I need (read: want) access to my files at all times (over a network is fine), especially since I maintain a library of what I've got on the TV computer. I don't want to have swappable USB drives, I want all hard drives available all the time on my network. I'm assuming that, since it's on a network, I won't need 16,000 RPM drives and thus I'm hoping a solution exists that can be moderately quiet and/or hidden away somewhere and still keep somewhat cool. So Slashdot, what have you done?"

Comments

Cheap NAS

[Bodhammer]

DLink - DNS-323 with two WD 1 TB Green Drives. Quiet, works out of the box and is also Linux hackable if you feel the need.
Enjoy!

My 2 cents

[ars vitae]

Personally I'm using a Synology solution at the moment, for my NAS. They offer a relatively low cost feature rich hardware, with low power and depending on the HDD's you use, lower power consumption than that of a always on PC.I've been thinking about later on upgrading, since a general rule of thumb, you can never have to much storage. For HD BluRay images I would recommend making sure the network isn't the bottleneck and use gigabit ethernet, as I'm finding on my aging 10/100 switches it's not cutting it. xvid's and MP3 streaming it seems to be fine.

Why do you need them available at all times?

[VinylRecords]

I used to work for ABC news and we never kept archive footage always accessible like you want. If we wanted something that was really old we'd have to dig it off a tape, an unplugged hard drive or powered off computer, or we'd have to find another news agency that had the footage and grab it off of a satellite feed. And this was a 24/7 TV news station responsible for national news programming where we would be tracking stories for years. If we didn't need a system where everything was instantly accessible then you needing it on an individual level might be overkill in my opinion.

I have over 30TB of music, movies, and raw video footage on my home computers and I just keep everything on separate external hard drives. I label the drives, back them up twice each, and then keep an index in a .txt file that is easy to search through. So if I want a 1080p backup copy of Blade Runner I search 'Blade Runner' in the .txt file and I see it's on drive 'A' and then I plug in drive 'A' and dump the movie on my computer. I also keep an external drive that has backups of every TV show I own on DVD. So if I want to watch The Wire then I plug in the external drive labeled 'TV' and have at it.

Re:Why do you need them available at all times?

[KahabutDieDrake]

First off, judging personal wants and needs by the way a giant corporation acts is hardly reasonable. ABC has cost/benefit to consider when trying to keep data available, and it's probably easier/cheaper to do it the way you say they do, rather than implement a fully digital, fully available storage system.

That being said, the solution is SIMPLE. If you have a bunch of hard drives with data you want, you put together a low end PC, install it into a server case, and fill it with hard drives and SATA controllers. When it's full, you build another one. You have 30tb of data, mostly not accessible. I have 10tb of data accessible from any internet connected computer on earth, and it's twice as much storage as I actually use. It cost me about 500$ to build and deploy a personal storage server, and it doubles as an HTPC. ( I already had most of the drives, and some parts) It's likely most people here have enough hardware laying around to implement a basic storage server. There really isn't any reason not to do it. As a bonus, since it's not a machine you need to access directly most of the time, you can hide it in a closet and forget all about it.

Sure, you could buy a premade NAS/SAN or stand alone data box. However, they are costly and not any more suited for the job than an old machine, or low end new system. At least, not in a personal environment. If you actually require robust data storage, I'd suggest a NAS, from any number of sources. But now we are talking about 4k worth of hardware, and requiring proper power systems to be added if you really want longevity out of it. However, that's overkill for a home storage solution, no matter how much data you have. Simply because you don't need enterprise class data serving, when only one or two computers are accessing the data.

If you don't know how to build and deploy a system with lots of drives accessible over a network, then you probably started at the wrong website for help. You want DELL/HP/IBM small office sales line.

Re:Why do you need them available at all times?

[gravis777]

For his "downloaded" 720p/1080p movies, its reasonable to assume that these are most likely reencoded to mp4 / mkv files or ts streams, probably between 2-15gig each. An external USB2 Harddrive should be able to keep up with the transfer rates. As such, you could probably go with something such as a USB hub and tons of external harddrives.

But I agree with you. I have DishNetwork DVR with the external Harddrive option. I currently have three external Hard Drives filled with movies. I keep a spreedsheet on the computer that tells me what I have on each drive. If you want to go fancy, CollectorZ has a great databasing program that you could probably use.

Now, as for his raw uncompressed footage, I have a couple of questions here - how many projects are you working on at once that you need to have tons of uncompressed footage, and do you really understand what Uncompressed means? Whether HDV, MP4, or MPEG2, every HD capture technology I am aware of uses some form of compression. The exception may be if you are scanning film in, taking several seconds to several minutes a frame, but it certainly does not sound like that is what you are doing.

I have been doing video editing for about 10 years, and always had space issues. I have handled this in two ways over the years. The first is to work on your project in 2-5 minute segments. Export those to an end format, such as MP4s, then stitch them together later.

Oh, but that's double recompression, you might say. Well, not necessaraly - you can easily use something LIKE VirtualDub (I dont think it currently supports MP4) to stitch video together without reencoding. This takes care of a LOT of harddrive issues when working on a large project.

The second is, and you won't like this, but hear me out, is to go ahead and compress your videos. I am talking about something along the lines of, oh, 20-40Mbps, depending on content, using a good codec and compression software. Saves a buttload of harddrive space, the drop in quality will hardly be noticable. Even then you can still edit, and export again in a fairly high quality. My experience is that you can probably go a couple of generations - IF you use a good codec, good software and high bandwidth - before you start really noticing degregation in video. Sure, if you compare the original to your finished work in front of a 30-40 inch monitor 18 inches from your face, you will probably notice a difference, but anyone else you show the finished product to - from a normal viewing distance, probably won't notice.

So to sum up, use external harddrives to put your "downloaded" content to, should be plenty of bandwidth on USB2 (or firewire800 if you are a stickler) to do it with, stop keeping around terrabytes of uncompressed video (any project exceeding, oh, lets use a magical number of 500 gig of raw source material, is insane), and, shoot, if all else fails, pickup a Blu-Ray burner. Discs are certainly coming down in price, although it is probably still cheaper to buy external harddrives.

Once again, as the parent pointed out, there is no reason AT ALL to keep all your videos online at once. Not even video production houses do that.

SATA port multipliers

[the eric conspiracy]

SATA port multipliers - 5 to 1 for about $50 + 5 2 TB gives you 10 TB off 1 SATA port.

Re:SATA port multipliers

[EdIII]

They work, but will slow the system down considerably. If you connect 5 drives up to 1 multiplier, the total speed you will get is the same as 1 drive hooked in directly. In otherwords, a bottle neck.

So technically it is possible to hook up 250 SATA drives into a single SATA RAID card, but you are not going to be that impressed with the performance.

Re:SATA port multipliers

[Kjella]

Well, from what I understood there are two modes - one which will give you only time slots so 5 drives each get 1/5th of the time. That's the cheap variety. The other variety is traffic based, you can't exceed 3 Gbps but you can get the cumulative read/write speed up to that point. The SATA spec site has more.

Re:SATA port multipliers

[gasgesgos]

That's only accurate under the assumption that a single drive can max out a 3 Gbps line. I'd like to see a reasonably-priced consumer grade HD that can pull THAT off. It doesn't really matter anyways as the ultimate bottleneck here will be the network at 1Gbps. Five drives evenly using a 3 Gbps channel would still be allowed 62.5 MB/s each, and that's still pretty good for network transfer.

WHS

[Barny]

Windows home server, 1TB 7200rpm main drive with seagate LP 5900rpm drives, lock it away and never have to think about it till you need to drop another drive in.

The reason for the fast main drive is that with WHS when you copy data to it, it stores it on the main drive first, then schedules it to be distributed out to the storage drives the next time a "storage balance" is done.

Works fairly well, its based off windows server 2003 at the moment, but if you can wait till the end of the year they have a server 2008r2 version coming out soonish.

Re:WHS

[Sabriel]

Another happy WHS owner here. I do recall reading that one of the service packs (there have been three) fixed the requirement for a big first drive - files now copy directly to the storage drives.

That said, I still use a fast system drive, and the rest are a mix of 7200 and 5400 rpm drives (depending on what was cheapest at the time).

Bought the original Coolermaster Stacker case. The front of the chassis is solely 5.25" drive bays - eleven of them - technically twelve if you mod the case to move the power+usb front panel elsewhere. :)

Oh, and despite being based on Server 2003, one of the nice things about WHS is that unlike the former it doesn't cost an arm and a leg.

Re:WHS

[Barny]

Yeah, because a bug (running vista on your client and using the server without the latest updates) over a year old and fixed is a problem...

I will also point out that the very first linux release wouldn't run on my 8088 cpu... ... ... ...

Please sir, if you are going to google for bugs, check your dates :)

Re:Something like this

[Anonymous Coward]

Do something like this. Put it in a case / box / cabinet of your own design since you don't need the rackmount capability.

http://blog.backblaze.com/2009/09/01/petabytes-on-a-budget-how-to-build-cheap-cloud-storage/

If possible use something like ZFS (or btrfs if you feel confident about it) so that you get checksumming data protection.

If you're going to put all your eggs in one basket, you better watch that basket very carefully.

The creators of that kit don't use any kind of redundancy with-in the box because their custom software stack handles replication (kind of like Google FS / Hadoop FS).

Re:Look at the DroboPro

[Firehed]

Having done that in the past, I'll say that buying a Drobo was worth the cost. Granted, I hunted around a bit to get a good sale price (it's not too difficult... though the FS is brand new so maybe not on that model yet), but unless you really enjoy tinkering with getting samba shares set up and working properly, sometimes it's just easier to buy your sanity.

Don't get me wrong - I wish they were cheaper. But their system worked better and more reliably than anything I ever put together, and I'm by no means incompetent. And their BeyondRaid tech, while proprietary, is pretty damn cool and works incredibly well. Being able to mix drives and not waste tons of storage space is a huge advantage that (as far as I know) I'm not going to get anywhere else.

Just a happy customer, not an employee or anything like that.

Two Options

[Zarjazz]

1) Cheap tower server + your favourite unix distro + software RAID + many, many cheap 2TB drives.

2) Standalone NAS device. Everyone so far seems to recommend different makes so I'll carry on the trend and suggest Thecus. Just slot in the drives and you're ready. Install the SSH module and you also have a Linux server too.

FreeNAS

[Anonymous Coward]

http://freenas.org/

Booting this off USB flash with 4 1.5 TB drives in a RAID 5 configuration =
It can even spin down the drives when they aren't in use.

Each 1.5 TB drive : $100
AMD Misc processor, motherboard with 4 SATA, GigE, RAM: $200
Case and Decent power supply: $100

Re:Define "massive"

[svirre]

My pricing indicate 2TB disks are slightly cheaper/GB than 1TB

Re:This should be modded up

[InsurgentGeek]

A contrary opinion. I have had a Drobo since the original release and it has been nothing but a disappointment. Drive incompatibilities, an extraordinarily high drive failure rate (at least 1/quarter)and a very confused partitioning scheme. Not something I'll repeat in the future. Oh, and data loss that had to be corrected via a firmware update. In short if I'm spending the money for Raid - I don't want to lose data. Period.

Re:Define "massive"

[Mr. DOS]

Check again: you're almost certainly comparing 1TB 7200RPM drives to 2TB 5900RPM drives. And Hitachi drives don't count, being the cheap pieces of garbage they are.

Cheap COTS NAS

[straponego]

I've been running off an Acer EasyStore H340 for about a month. I'm very happy with it; it's very cool, quiet (if anything else is making noise, you won't hear it; in a closet, you definitely won't hear it) and plenty fast for most households. 4 hotswap SATA bays, eSATA, USB, and GigE. I can push 75MB/s via NFS to and from it (reading and writing from RAM), which is plenty for streaming video. It comes with Windows Home Server and it's headless, but I popped the drive into another box and installed Ubuntu with an SSH server. Worked like a charm. I'm also running a 2TB software RAID1, mt-daapd (iTunes) and squeezebox servers. I'll probably put Samba on it too.

The only thing I'd change is that a dual core Atom would be better. I actually haven't run into a bottleneck yet, but I wouldn't try reencoding videos on it. I believe the dual core model will be out this month. No affiliation with Acer; I'm just geeked because this is just the quiet, cheap server I've wanted for years. Sounds like sharing your other computers via NFS (automount) or CIFS plus one of these would address your needs; if not, maybe the info will help somebody else.

ZFS + Ubuntu

[Anonymous Coward]

I wanted to do a similar task myself this is my solution from 6 months ago:

total cost under $1000, more than half of that was for the harddrives

mobo with lots of SATA ports (6+)
low power AMD CPU (but make sure it scales up, ZFS uses a lot of CPU cycles under use, mine is 1.0Ghz to 2.4Ghz dual core, but only 20-40 Watts)
1 Tb Hitachi Drives to fill it with (surprisingly seem the be the most stable option these days)
1 old hard drive to use Ubuntu and software with (I use a 60 Gb)
Ubuntu and zfs-fuse.
netatalk and avahi for Mac access, also use ssh and svn for important file storage and access, but I'm a command-line junkie.
low power system (less than 100 watts), very low noise (invest in good CPU/case fans)

To save room in my small apartment, I also use this as my main desktop, upgrading my previous 7 year old computer.

Notes on ZFS:
Amazingly easy to use and expand data with. Great backup features. Stress free storage.
Very slow though with database and random access files (photo management/websites/etc),
so I recommend you use your active copy of these on the standard drive.

Re:Define "massive"

[ls671]

NTFS algorithm is kind of silly and will pick the first free block it encounters to begin to write a big file *without regards* for the amount of consecutive free blocks available there.

On the Windows install you are talking about, all it takes is a little temporary file getting deleted. NTFS encounter a couple free blocks freed by that file and starts writing there only to realize there is not enough space for the whole new file it needs to write once it uses the last consecutive free block available there.

In contrast, even ext2 FS keeps some knowledge of the global state of the FS so it will pick the best free available block to start to write the file *with regards* for the amount of consecutive blocks available.

I have never defragmented an ext2 FS in my life, the write algorithm takes care of keeping fragmentation low.

Re:Something like this

[mindstrm]

Unless they changed something since they published their howto - your BZZT has it's facts wrong....

Backblaze pods use 45 drives in each pod.

Each pod is made up of 3 Raid-6 volumes.

Each Raid-6 volume is made up of 15 drives (13 drives + 2 parity)

There are two parity disks for every 13 data disks. That's higher than you might want to go in a normal enterprise setting, but as they also handle redundancy between pods, it's an acceptable tradeoff in their case that I'm sure they've calcluated out quite well given drive sizes and rebuild times and all that......

Re:Look at the DroboPro

[AmaDaden]

I have a Drobo and a DroboShare. The DroboShare runs a slimmed down version of Linux so a network attached Drobo uses typically uses samba. The benefit of having one is NOT the ease of software set up. The reason I love it is the ease of drive management and small hardware size.

Due to the small size and slick style I keep mine in my TV cabinet. I've done the measurements and no PC case on newegg can fit in this same space, never mind something that can house 4 harddrives.

The other thing that is so valuable about a Drobo is how well it manages it's RAID array. They call it BeyondRaid but I hear it's just a as many normal RAID arrays as it needs to organize the drives to both optimize space and maintain redundancy. Also you can pop harddrives in or out while it's on and it will automatically restructure the RAID on the remaining drives to still be redundant with out any need to shutdown or stop sharing data. I recently needed to test this out for my self. I popped out my 4th drive, plugged it in to my PC, formatted it and started moving data from my Drobo to the harddrive I just removed from it while the Drobo was still restructuring. I expected a huge mess, but everything worked exactly like the advertised. I was kinda shocked.

FYI the reason I did that swap out was because I foolishly formatted my Drobo as NTFS. This worked ok but I had one to many problems talking to it from my Linux PC. The permissions were all messed up over samba. New folders and files I created on the Drobo were root access only for some weird reason. So I decided to format it as ext3. Since the DroboShare runs Linux this is the best option for a shared drive and works fine while talking to mac and windows as long as you do so over the network.

RAID 5 or 6

[Hamsterdan]

Creating a RAID 5 array on OS X, is really not that hard. On Windows, pretty trivial too, never got around playing with it on Linux though, must be pretty straightforward.

Personnally, I prefer SOFTWARE RAID, because it's easier to recover from a failed array compared to a hardware card that is not manufactured anymore.

(example, take the drives to another MAC or Windows machine, they will be recognised automagically if it's done with the OS, not so if the controller went belly up...)

I'm running a raid 1 on an old G4, and RAID-5 on an old Win2k Server machine, and they are easier to manage than the one in my Proliant (Hardware)

Re:Define "massive"

[inKubus]

What you want is cheap 5U rack servers with either OpenFiler or FreeNAS. Personally, I like openfiler better. iSCSI is going to be the way to go unless you want a thick OS on the server and all the other admin issues that come with that. Plus, with openfiler you can still do block level snapshotting and change replication. Also, I've heard good things about Open-e as well. And if you want to mess with ZFS, there's OpenSolaris.

What you do is get yourself a huge (4 or 5U) barebones server from newegg or a cheaper place. Make sure to get a couple of good SATA RAID controllers. Not FakeRAID! SAS would be better, but the drives are a lot more, even for the nearline drives that are basically SATA drives with a SAS interface. Adaptec makes some real SATA raid cards, and there's 3Ware as well. You don't have to worry a lot about the cache, but if it isn't battery backed, you're going to write though it anyway. Who cares, you have 16 spindles! Load it with a bunch of drives. They don't have to be the biggest, anyway more spindles means more performance. 16 500GB drives is going to be fine, for instance, because then you can take 1/3 of that for RAID 6, have some hot spares, etc. Get the slowest drives you can, maybe get a little SSD to use as a boot drive (there are small ones for around $100). You could even boot from a USB key if you feel like the hassle. You don't need a ton of processor. A celeron would probably work, but you probably do want something 64 bit so you can put a bunch of RAM in it as you get more advanced.

Also check out Storage Search. Not a very well designed site but tons of goof info under iSCSI and SAN and NAS.. If you're rich, you might try out an EqualLogic, they are around $28,000 for 8TB but pretty slick.

Re:Define "massive"

[Anonymous Coward]

Except that this is completely untrue. Even Windows 95 scanned the table of free blocks for a reasonable area of consecutive free blocks.

  HKEY_LOCAL_MACHINE \System \CurrentControlSet \Control \FileSystem
DWORD ContigFileAllocSize

http://technet.microsoft.com/en-us/library/cc768196.aspx

Re:Define "massive"

[BikeHelmet]

When I hear a question like this, I usually recommend heading over to the NCIX forums. There's some crazy guy over there - death_hawk - building a 100TB array.

What I did was a bit less ambitious. A regular old NAS running off a cheap non-RAID SATA card in a case with lots of HDD bays.

For interest, I'll throw up a build that easily scales to 12TB. Since you mentioned noise, I'll prioritize that instead of capacity. I'll use a case geared for silence, a fanless mobo/cpu, a quiet PSU, WD Green HDDs, and a ridiculously cheap SATA card.

Case - 8 bays: http://www.ncix.com/products/?sku=51277&vpn=6900654&manufacture=Fractal%20Design *1
Motherboard/CPU - Silent: http://www.ncix.com/products/?sku=50891&vpn=AT5NM10-I&manufacture=ASUS *2
DDR2 - 1GB: http://www.ncix.com/products/index.php?sku=18584&vpn=VS1GB667D2&manufacture=Corsair&promoid=1114 *3
PSU: http://ncix.com/products/?sku=33357&vpn=CMPSU-400CX&manufacture=Corsair&promoid=1114 *4
SATA Card: http://ncix.com/products/?sku=19892&vpn=SY-SA3114-4R&manufacture=Syba *5

HDD - 2TB 4KB http://ncix.com/products/index.php?sku=49591&vpn=WD20EARS&manufacture=Western%20Digital%20WD&promoid=1114 *6
HDD - 2TB 512b: http://ncix.com/products/index.php?sku=36130&vpn=WD20EADS&manufacture=Western%20Digital%20WD&promoid=1114 *7

OS: FreeNAS, Ubuntu, Win7, Other *8

*1 Only six will be filled. 6 SATA ports.
*2 Case still requires fans/airflow.
*3 A NAS probably only needs 512MB, but 1GB is cheap. A Win7 NAS may benefit from 2GB.
*4 Must be capable of spinning up 6-8 HDDs at once.
*5 Must be flashed with new non-RAID BIOS to avoid silent data corruption for > 1.0TB HDDs; disk read/write speeds around 30MB/sec, in my experience, on ext2. (but running with a VIA CPU - not dual-core Atom)
*6 Must be specially formatted under Windows and Linux. (Most distros only support 4KB sectors when the drive reports 4KB - these report 512b to maintain XP compatibility)
*7 May have longevity issues. (too early to say right now - lots of complainers, which reminds me of the 7200.10 days. A heck of a lot of those chirping barracudas perished early)
*8 Please verify SATA card support first. Ubuntu and FreeNAS work fine with this card, but I've never checked if Win7 has drivers. Do note that you'll have to flash it. *9 If that's a problem, buy a more expensive card. (which may give better performance, and SATA2 support) Promise makes nice non-RAID SATA cards.

*9 Flashing the PCI SATA card requires making a DOS boot CD: http://www.hiren.info/pages/bootablecd

Please note: A solution like this will take 12+ hours to set up. It's highly likely you'll blow a whole weekend, even if you know what you're doing. You may have to try multiple distros to get proper Atom D510 support, unless you go with Windows. When I put mine together, atoms weren't available affordably, so I went with a cheap VIA board. Ironically, Ubu

Re:Define "massive"

[jridley]

I have a lot of storage for the same reason that the OP does, and I PREFER 5400 RPM drives. They run cooler and are still faster than what I need.
I prefer WD Greens, but Samsung EcoGreen works well too. I buy the green ones because, again, they run cooler.

1.5TB drives have been cheapest $/GB for a while now, though I suspect 2TB will take its place, especially after the 3TB drives hit the shelves.

42TB for under $4k

[DedTV]

Here's my setup that's currently capable of holding 44TB of storage (I have 18TB so far). Nothing fancy, Just something to hold all my media that isn't horribly noisy or hot and that was still relatively cheap. I have it sitting on a coffee table in my home office so you could put it pretty much anywhere.

A $320 Norco 4020 case that has 20 hot swappable drive bays plus 2 more fixed drive bays inside. http://www.newegg.com/Product/Product.aspx?Item=N82E16811219021
A $250 server motherboard with at least 2 PCI-X slots. I chose http://www.newegg.com/Product/Product.aspx?Item=N82E16813182142 because it takes a Core 2 CPU and DDR2 and I already had plenty of those laying around so I saved a few bucks in parts. I also had a CORSAIR CMPSU-850HX laying around and used it for a power supply which runs about $190.
$99 SUPERMICRO AOC-SAT2-MV8 SATA card. 2 of them will fill the server if you pull the DVD drive (I use an external anyway). 6 sata on the Mobo + 16 more from the 2 cards. Some people complain they're slow but I can pull 60+ Mb/s over the network from them. My guess is they're putting them in regular PCI slots on regular MoBos and not PCI-X slots on a server board.
For an OS, I simply use Windows Home Server. It's $99, windows simple, and is perfect for just storing video files. Reinstalling the OS can be a massive pain though as WHS reinstall script thing never works when there's a controller that WHS doesn't support out of the box (ie. the Supermicro cards). And the new version of WHS based on WS 2008R2 is on the way and there won't be an easy way to migrate.
I also use Flex Raid (Software Raid 4) and sacrifice one disk as a parity drive because duplication isn't much safer but eats a lot more of my space. I just have it do the rSync when no one is likely to be doing anything with the server so it's never a hassle.

So the base cost is within a few hundred bucks on either side of a grand. Less, if you have parts that can be cannibalized from old machines

From there, I just add Western Digital Caviar Green 2TB drives as needed. They're cheap, quiet and run cool. The EARS ones need a jumper (and none are included with the drive) to run under WHS but are $15 cheaper than the EADS ones on Newegg. And of course you can use any old drives you have laying around too.

I have Acer Aspire Revos ($330 http://www.newegg.com/Product/Product.aspx?Item=N82E16883103235) running XBMC Live installed on a USB Drive (xbmc.org) hooked up to each of the TVs in the house and use wired Gigabit (provided by whatever the cheapest 5 port Gigabit switch was on Newegg at the time, I think it was about $25) to stream DVD ISOs from the server to the TV. I don't have much HD stuff and the Revo can get bogged down when you try to play really high def video. 10Mbps works fine with XBMC using vdpau, but an 18Mbps MKV was a bit too much and went slideshow in places.

Best cost/benefit, if your desktop runs Linux...

[durval]

... is to simply put the extra disks on it.

My main desktop at home has been running Ubuntu Linux for the last 6+ years, and for the last 3 years I migrated the functions of my NAS (which used to run on a separate linux box) to it, so I have one box less to administer, to consume power, or to break down.

I have currently 10TB of total disk space using 5 2TB disks (8TB usable, when you account for the RAID5 redundancy overhead), but could easily migrate to 20TB on 10 2TB disks (16TB usable, and gaining extra redundancy by moving from RAID5 to RAID6). The result is:
- Very upgradeable (starting with 250GB disks back in 2004, I've migrated all the way to 2TB disks, and will continue doing so; the old disks are simply replaced with the newer/bigger ones and re-purposed as off-line storage, being plugged on a eSATA dock when I need them
- very fast (as the disks are on the machine itself, it's much faster than acessing the files on a NAS over the network);
- very usable (it's all mapped on a couple of ReiserFS filesystems, created on top of LVM volumes, directly accessible without needing to mount anything or configure anything over the network);
- very reliable (I'm protected against any one of the five disks failing, thanks to RAID5 configured on top of Linux MD, through for more disks RAID6 is really recommended, and ReiserFS in my experience is very reliable against crashes and power outages, at least *much* more reliable than EXT3 and XFS).
- very cheap: I've used the SATA controller already available on my motherboard, providing for 6 SATA disks, and apart from the disks I only had to spend money on a multi-disk internal rack: these are great, they fit 5 SATA disks on 3 x 5.25" bays on the front of your desktop, are very cheap (around $75) and give you hot-swapping and great ventilation (via a large, low-noise fan in the back) to boot. Just be sure to use a computer case that has no "rails" or other protuberances between each 5.25" bay, or else you won't be able to insert the rack as it spans 3 bays.

You could use a SATA RAID controller (or even SAS disks and a SAS controller), but I found that it's quite expensive, and unnecessary as the above setup gave me all the speed I needed, and them some.

In short, I'm very satisfied with my setup, and I recommend it to anyone who has large disk space requirement at home.

Some pointers to the hardware I'm using:
- Motherboard: Asus M4A78-EM which is reasonably cheap, very stable and has 6 SATA ports (5 internal and 1 external), fitting the bill perfectly;
- Disks: Seagate SATA 2TB 5900RPM Retail kits : The retail kit (instead of the bare OEM drive) gets you a disk with FIVE years of warranty (instead of just 3 years) and comes much better packaged (so reducing the chances of early death due to shocks during transportation).
- 5-disks-on-3-bays internal SATA enclosure: NORCO SS-500 : great little bay, as described above. - External eSata dock: Startech SATADOCKU2E: with it, when I replace my old (smaller) disks with new big ones, I can re-purpose the old ones immediately as off-line storage,very efficiently (my motherboard already has an eSata connector) and very cheaply (I store the disks on plastic storage cases when they are not docked, very cheap and compact.

Hope the above is of help.

Re:This should be modded up

[gbr]

I got a Drobo as well, the 4 drive USB/Firewire version.

I had a RAID array die, and I needed something fast to save my data before the weekend was over. The Drobo worked.

I'm now back on a Linux RAID system, and the Drobo is relegated to backups. Why? It's slow. Way slow. I wouldn't recommend for daily use.

Use D-Link Drive Enclosures - we do :-)

[code-dweller]

We have a similar problem at the Mad Lab - we generate a lot of data (from our digital studio and other projects), we need access all the time, and we need reliable storage.

At first we were putting lots of drives into a PC -- but that led to problems. For one thing there was a single point of failure (main board, power supply, take your pick). Another problem was that the system was loud and power hungry. Then there was the backup problem -- there was no efficient way to do it without building another system just like it --- you can't ship TB of data off-site via the 'Net for backups, it just isn't practical. Then to make matters worse we decided we couldn't do anything else with the server without putting our data at risk... that was the last straw for me -- The server was overkill for the task and couldn't be used for anything else. I was stuck in a paradigm - I knew better - but I'd forgotten that temporarily...

Then I hit upon the solution of using D-Link 2-Bay network storage devices. http://www.dlink.com/products/?pid=509 These little guys are reliable, solid, efficient, and affordable. They're also pretty green because they will spin down when you're not using them thus saving power and POH time on the drives.

We use them in pairs: NS0 for storage, and BS0 for backup. With a firmware upgrade they will do NFS - so we have one of our servers map the two devices and then rsync NS0 to BS0 once per day. Newer versions may have the NFS capability built in (it was on it's way, it was beta firmware when we did it).

Now we have the key features of the high-end NAS solutions we use in data centers, but we have it on the cheap. The solution is scalable (more storage, more enclosures), reliable (mirrored drives all around, fast and easy to access (NFS or Samba - take your pick), provides redundancy (outboard power supply for each enclosure - easy to swap, separate controller for each pair of drives), and easy to manage (what's not to love about a scheduled rsync task via nfs for automated backups?).

We can easily access the data from either windows or *nix boxen on the internal network without any trouble. When we need to access the data from outside the Mad Lab we shell into a server and sftp what we need from there.

Here is a pic of the two of them in the rack: http://www.lifeatwarp9.com/wp-content/uploads/2010/01/MadRackBefore-225x300.jpg We've been using this setup for quite a while now and it's been dead solid. When we need to expand we just plug in another unit and map it. We've also installed this kind of configuration in customer facilities to manage backups and solve other storage problems on the cheap.

_M

Re:Define "massive"

[schon]

When a drive fails, SW RAID doesn't allow one to boot the system unless one has (1) Used RAID-1

True enough, but (honestly) how hard is it to use RAID-1 with hot spare(s) for your boot partition, and RAID5/6 for everything else? (answer: not very, I'm doing this at home.)

Setup the BIOS to use the two mirrored disks successively while booting.

Most modern motherboards already do this. Contraiwise, even if you have to go out of your way, it's *still* much easier than screwing around with driver disks for HW cards when installing.

SW RAID often exhibits very poor performance when a drive fails as the underlying drivers want to try to make the operations against the disks work and will often expend a fair amount of effort to retry operations and wait for extended periods of time to force the disk operation to work.

Never seen this happen. "bad" drives on SW RAID mark out just as quickly as those on cheap HW controllers.

Standard disk controllers do not support hot swap. So when a drive fails, replacing the drive involves shutting down the server, swapping the drives, and then bringing it back up.

Bull-fucking-shit.

I regularly attach and detach drives while my home server is running (I back up to external HDs, connected to the internal disk controller with an e-SATA cable.) Motherboard is an ASUS P5EVM-DO with a "standard disk controller".

Re:Define "massive"

[PipsqueakOnAP133]

I'm not sure even more than 1 core is necessary for software RAID.

I just (last week) built a OpenSolaris based ZFS NAS with 2 drives on a old Core Duo. Its never gone past 20% CPU utilization even when I'm copying from 3 disks over gigE to it with on-the-fly gzip compression enabled as part of ZFS. I think it sustained 48MB/sec.

CPU would have been a Core Solo instead had the ebay seller not mailed me the wrong chip.

I'd say anything P3 class or better would be fine with one core.

Also, I'd object to saying that OpenSolaris is a bad choice. We're taking file servers here. ZFS works well and OpenSolaris is open source. It's not like you want bleeding edge commits from a tree to run the thing that stores your precious data anyways.

Finally, if this is dedicated to being a file server, might as well boot off the USB flash drive or liveCD. Your distro ought to be small enough to fit and you have less chance of OS corruption from a read-only image.