Comparison of Nine SATA RAID 5 Adapters

Robbedoeske writes "Tweakers.net has put online a comparison of nine Serial ATA RAID 5 adapters. Can the establishment counter the attack of the newcomers? Which of the contestants delivers the best performance, offers the best value for money and has the best featureset?"

Comments

Eight or Nine?

[SirTwitchALot]

TFA says nine adapters, but the graphic says eight, whoops.

Re:Eight or Nine?

[pbranes]

They must have started counting at 0. Stupid off-by-1 errors. ;-)

Seriously, though, I have been seeing many servers start to come in with SATA drives. Right now it is low end and off-brand servers. Dell even ships SATA drives in their cheapest server line. Sure SCSI has high spin rates & throughput, but they are freakin expensive. A good SCSI raid controller costs close to $1000 and a good SCSI hard drive can cost $400. It is so expensive, that it is reallly worth it sometimes to get the SATA drives in servers. I haven't seen that reliability of SATA over SCSI is a problem. I'm truly hoping that SCSI goes the way of the dodo. Its a pain to use. Who know what kind of cable you're supposed to use with that external SCSI device. SCSI, in its current form, is just opening itself up to becoming antiquated.

Re:Eight or Nine?

[ackthpt]

TFA says nine adapters, but the graphic says eight, whoops.

It was a parity bit, ignore it.

Re:Eight or Nine?

[Rei]

Yeah, but what sort of real-world operating performance improvement will you get with SATA? I can understand using non-scsi raid to add redundancy, but to improve performance seems kind of silly. In the real world, your delays on the vast majority of files are not in throughput, but instead in seek time and latency. And as far as these things go, even the best ide drives are pretty bad compared to even moderate performance scsi drives.

If you just want redundancy, go ahead. But if you want better system performance, you need better seek time and latency. Keep your root partition on a small, fast scsi drive, and use cheap storage for your "bulk" content.

Re:Eight or Nine?

[Phantom69]

From Page 2 of TFA:

Note: Since the original Dutch article was published in late January, we have finished tests of the 16-port Areca ARC-1160 using 128MB, 512MB and 1GB cache configurations and RAID 5 arrays of up to 12 drives. The ARC-1160 was using the latest 1.35 beta firmware. Furthermore, a non-disclosure agreement on the LSI MegaRAID SCSI 320-2E PCI Express x8 SCSI RAID adapter was lifted. The performance graphs have been updated to include the Areca ARC-1160 and LSI MegaRAID SCSI 320-2E results. Discussions of the results have not been updated, however. The results should be self-explanatory.

Re:Eight or Nine?

[FemmeT]

There are really nine adapters: 3ware Escalade 8506-8, 3ware Escalade 9500S-8, Areca ARC-1120, Areca ARC-1160, HighPoint RocketRAID 1820A, LSI MegaRAID SATA 150-4, LSI MegaRAID SATA 150-6, Promise FastTrak S150 SX4 and RAIDCore BC4852.

The results of the LSI MegaRAID SATA 150-4 and MegaRAID SATA 150-6 have been combined in the graphs since there is basicly no performance difference between to two in configurations up to four drives.

Re:Eight or Nine?

[operagost]

Who know what kind of cable you're supposed to use with that external SCSI device.

RTFM. And the answer with anything made in the last three years is "68-pin LVD". I suppose there might be some SE and HVD stuff still around, but you can tell just by looking at the symbol by the connector.

SCSI has other advantages, such as the ability to have fifteen devices per port. I would love to see that SATA raid card with 15 ports stuffed into it. Plus, your cable complaint is also an issue with SATA, as there are some manufacturers using the flimsy internal cable on external drives in addition to the new SATA II standard.

Re:Eight or Nine?

[Donny Smith]

> I haven't seen that reliability of SATA over SCSI is a problem.

And did you know that rebuilding a SATA RAID can take ages (which overlaps with times when you need every bit of performance you can get)?

> Who know (sic!) what kind of cable you're supposed to use with that external SCSI device.

How many companies have a single disk or JBOD (or RAID without enclosure) on an external SCSI connection without enclosure?

> SCSI, in its current form, is just opening itself up to becoming antiquated.

Everything it its current form is, including SATA.
Imagine how stupid would it be to still have SATA storage in 2015? I hope disk drives and all current interfaces will be out of fashion...

Re:Eight or Nine?

[Tassach]

SCSI, in its current form, is just opening itself up to becoming antiquated

It would be more accurate to say that the SCSI pricing model is becoming antiquated. Vendors have gotten used to being able to charge a 300%+ premium for SCSI hardware because, until recently, it was the only game in town for serious server storage.

The current generation of SATA gives you roughly 90% of the performance of SCSI for less than 50% of the price. Unless you absolutely need every shred of I/O throughput money can buy, the price/performance ratio of SATA makes it an attractive and realistic alternative.

Re:Eight or Nine?

[Aroma 7herapy]

and use cheap storage for your "bulk" content

erm, perhaps you should try and remember what the acronym RAID stands for...

It's just moved on to the point where SCSI is the high-end solution and SATA the inexpensive (hint) one.

Re:Eight or Nine?

[Woody77]

SATA2 is allowing port multipliers. Which allows a fair number (possibly 15, don't remember) of drives per port on the controller.

But you get the same problem you do with SCSI. Sure, you can hang 15 devices on that 320MBps bus, but you can only push 320MBps of data through it. With SATA (and SAS) you get the full 150MBps per drive, which means that when you saturate your SCSI 320 bus at 3-5 drives or so, a 6- or 8-port controller can keep scaling up in bandwidth.

Add in the new per-port throughput specs for SATA2, and the port multipliers, and you've got a LOT more bandwidth than SCSI has ever had (aside from fibre, which is a whole different beast).

Re:Eight or Nine?

[Anonymous Coward]

SATA can effectively called SCSI. Aside from the serial interface, there is little difference. For entry-level NAS servers, SATA and a good RAID backbone works very well even with heavy LAN saturation and complex RAID table calculations. If you devote, say, a Pentium 4 system to only NAS, software RAID using Linux creates a much broader range of possibilities that rival enterprise-class systems at a fraction of the cost.

Re:Eight or Nine?

[Kent Recal]

A good SCSI raid controller costs close to $1000 and a good SCSI hard drive can cost $400.

Depends on size and features. Our Adaptec U320 controllers cost $150. 36G fujitsu SCA drives cost $100 a shot. Speed, reliability and hot-swap capability are well worth the money when your server is actually doing something worthwhile...

Re:Eight or Nine?

[bondjamesbond]

I agree that SCSI is much more expensive than the SATA stuff. So, we have a new standard: RAID and RAED (Inexpensive vs. Expensive). Pronounced "RADE" and "RED".

Re:Eight or Nine?

[Anonymous Coward]

"It was a parity bit, ignore it."

Ignore parity bits? That's crazy!

Re:Eight or Nine?

[hey!]

No, I think we're suppose to reconstruct the ninth view from the other eight.

Re:Eight or Nine?

[Cylix]

They did miss a good feature of the LSI MegaRaid.

It supports a daughter board for controller redudancy.

Note, I don't have the daughter board and I can't test how well it works. Overall, I think the feature set was a bit understated. (It's definately in the affordable range too)

Though I would have liked a non-host based option for raid access, it was one of my more appealing choices when I had a new system come in for backups.

32 pages? No thanks.

[Mr. Sketch]

After 32 pages, it's probably just best to skip to the conclusion:
http://www.tweakers.net/reviews/557/32

Where it has the executive summary:

Areca ARC-1120: highly recommended
RAIDCore BC4852: recommended
HighPoint RocketRAID 1820A: recommended

For several reasons, we will refuse recommendations on the remaing adapters in this comparison

I think that pretty much covers the jist of the article.

Interesting that the 3ware offerings performed...

[tabkey12]

quite badly! They've been synonymous with quality in the RAID industry for many years. Look at this:

3ware Escalade 8506-8 is lagging far behind the competition. Moreover, it misses important features such as online capacity expansion, online RAID level migration and RAID 50 support.

http://www.tweakers.net/reviews/557/6

What they say in the article is almost damning really...

3ware me

[nfsilkey]

In most aspects of my life, I try to not be a snobby douche. But when it comes to hardware EIDE/SATA RAID, why would you think of anything _but_ 3Ware? Theyre the kings and have been.

Re:3ware me

[tabkey12]

Because of this perhaps. Occasionally it helps to read an article before making comments that just make you look uninformed.

Re:3ware me

[B3ryllium]

The previous poster had this to say:

3ware Escalade 8506-8 is lagging far behind the competition. Moreover, it misses important features such as online capacity expansion, online RAID level migration and RAID 50 support.

Re:3ware me

[B3ryllium]

Way to make me doubly-redundant :) I wanted to post A/C, but apparently my subnet has been abusing AC privileges lately so I'm not allowed to.

I guess that means someone in my neighborhood (near Mt Doug Park) is a slashdot troll.

Re:3ware me

[Anonymous Coward]

sorry, but I'd still rather be backed up than knocked down! Anyway, you got a higher mod than me, so don't complain :)

Re:32 pages? No thanks.

[tabkey12]

But RAID is nota one size fits all game - the detail of the article is extremely useful for people who will be tailoring their RAID to a specific application. Yes, this article is specialised, but I hardly see how reducing it to a list of three, relatively meaningless names is helping.

Don't plan on mixing Highpoint cards

[GoodNicsTken]

I had a Rocket Raid 100 (IDE 4 drive RAID1/0) and a RocketRaid 1640 (4 Channel SATA RAID 0,1,5) card. With nothing connected to the 1640 and 2 mirrored drives on the RR 100 the disks attached to the RR100 in bios show up on the 1640, and when windows gets to the boot screen it locks up.

When I removed the drives in windows, it booted up without problems. Highpoint has sent me diag tools to run rather than building this in their lab!

I'm not too impressed with them so far.

Re:Don't plan on mixing Highpoint cards

[ackthpt]

and when windows gets to the boot screen it locks up.

When I removed the drives in windows, it booted up without problems. Highpoint has sent me diag tools to run rather than building this in their lab!

I'm not too impressed with them so far.

Perhaps it's just another Windows bug? There's scads of them, you know. I remember wasting a couple days trying to get two controllers to behave, only to find windows identified them wrongly and created a conflict. It was a while back I don't recall all the details, but it was one of those deals where Microsoft Tech Support actually suggested re-installing Windows. Which must be a stock answer for everything. Eventually I was elevated to talk to someone with a brain and they explained what went wrong and how to fix it. Not exactly what you're going to find in an FAQ anywhere.

Re:Don't plan on mixing Highpoint cards

[Nik13]

I've used a LOT of different IDE RAID cards (promise, sil image based, HighPoint based, ... most of the stuff around).

The only one that ever gave me a problem of all of them was a Rocket Raid 100 (HPT370A based). Promise has the best card of them all IMHO (performance wise, good drivers, ... good overall). The sil image stuff is generic but it's very inexpensive and works surprisingly well (slow in DOS compared to promise cards though). Nice thing about them is most of the time they have a jumper to make them either RAID or a plain IDE controller to fit 4 more drives. (the promise has a hidden resistor hidden under a IC socket and needs you to reflash). I like the flexibility, it can be reused for different purposes in a different PC later on. The HP never amazed me. Not the best performance, not the best price, no particularly great drivers... And the only card that gave me problems was based on this. Eventually on one of them a port (one of the data pins) died. Hopefully their SATA stuff is better (but I tend to attribute my problems more to QC on this line of cards more than anything). Given another alternative, I'll stay away from HP, just like I stay away from VIA since their dreaded KT133 chipsets.

Re:Don't plan on mixing Highpoint cards

[reassor]

I've used a LOT of different IDE RAID cards (

I have had the luck to use a 3Ware 2Port IDE-Raid-1 Card some weeks ago.As a mid-educated-Computer User,i had no Problems to create and use the Raid1-Array.A few keystrokes there,a few clicks somehwere else and i was ready to install Windows 2000 on the 2 Drives.

So now Raid-Solutions might not be complicated,was my first thought.After klicking to the 32Page-Raid "Guide" and reading this Slashdot-Topic,i thought wrong.Many People are struggling without knowlege in deep waters.They pay high Amounts of Dollars,only to find out,that Raid-Controller "X" is not supported in "Linux.XY" or that some (serios) Bugs make the huge Pile of "highend-highmoney-Waste" unusable.For them,i think,a Arena-Raid-Box (these Black Boxes with a Ethernet-Port,which provide Raid to someone) was better.

I hope or i pray,that nobody just buy this,because it looks so good...I hope,they have a serios Use for it.

Enjoying this

[Killjoy_NL]

When I see a link to such a great and extensive review and it's also from a small dutch tech news website, I feel proud to be dutch :)

Re:Enjoying this

[Anonymous Coward]

I feel proud to be dutch :)

There are only two things I can't stand in this world. People who are intolerant of other people's cultures... and the Dutch.

I dread to think

[Linker3000]

I dread to think where my 'vanilla' dual channel SATA controller would come on the evaluation list but, hey, it's working fine and only cost £25!!!

Re:I dread to think

[ackthpt]

I dread to think where my 'vanilla' dual channel SATA controller would come on the evaluation list but, hey, it's working fine and only cost £25!!!

That was hardly insightful. Maybe a couple more exclamation marks...

How about a link, what you have hooked up to it, what kind of load it experiences, how you use it, any fiddly bits concerning configuration -- all would be helpful.

I've got 2 160GB drives sitting on the floor and haven't done bugger with them since getting them months ago, trying to decide how I want to deploy them. (It's nice having a connection at a drive manufacturer and getting them at a discount, but you still have to figure out what you're going to do with all that space.)

Re:I dread to think

[mrhight]

send 'em my way...

Re:I dread to think

[Linker3000]

It's One of these (oh, yeah I bought two for £25!), hooked up to two of these in an Acer Altos G310 P4-2.4 with 768MB RAM running Centos-3.

The system is running eGroupWare for around 40 users and is also a store for their mailboxes. Load is not that heavy and such a non-issue that I've not bothered to benchmark anything

There was no hassle installing the drivers from the manufacturer's Web site. The initial RAID 1 sync on the disks took 90 mins.

Re:I dread to think

[Anonymous Coward]

There's an excellent scene in the Krzysztof Kieslowski file Three Colors White where the protagonist Karol Karol is attempting to buy a plot of land from an elderly and mulish Polish farmer. Sensing that the sale would mean a lot to Karol and having no use for the money, the farmer is reluctant to sell.

Karol attempts to seduce the farmer with several ideas for spending the money but there appears to be deadlock. Until an avaricious smile spreads slowly across the farmer's face and he says:

Yes... I could bury it in a hole.

Jesus, use the disks for something.

Re:I dread to think

[chthon]

• Create User Mode Linux Instances
• Keep a mirror of Debian sarge and sid
• Keep several kernel source trees for experimenting
• Keep backups of your data
• Try to use the system for testing Linux enterprise features, but never find the time

Just some of the things I am doing with my system (which has BTW, also 2 Gig RAM).

Re:Comparison of Nine SATA RAID 5 Adapters

[killmenow]

*cough* TROLL *cough*

You know the cheap-reliable-fast triangle.

[lanc]

Well, cheap+reliable == linux + softraid + Enhanced Network Block Device + Enterprise Volume Management System (or LVM2). It is often faster than non-hw-raid (fake-hw controllers.

Re:You know the cheap-reliable-fast triangle.

[shic]

The cheap/reliable vertex of cheap; reliable; fast triangle is fairly obvious.

Can you suggest such a clear cut answer for the "cheap, fast" vertex?

[Yes - I would have a use. No - this wouldn't mean I'm playing fast-and-loose with unrecoverable data!]

Re:You know the cheap-reliable-fast triangle.

[Kent Recal]

Linux softraid is fine (stick with raid1, though...) but ENB and LVM are a nono if you need it reliable. Do not try this on a prod. system unless you intend to go through hellfire, soon.

Re:You know the cheap-reliable-fast triangle.

[the eric conspiracy]

Can you suggest such a clear cut answer for the "cheap, fast" vertex?

How about /dev/null ?

Re:You know the cheap-reliable-fast triangle.

[shic]

Can you suggest such a clear cut answer for the "cheap, fast" vertex?

How about /dev/null ?

Ha-ha. But... I find /dev/null to be extremely reliable - it is the most consistent of any device I to which I have access.

Re:You know the cheap-reliable-fast triangle.

[lanc]

it's obviuos. softraid, RAID0 (striping).

Re:Interesting that the 3ware offerings performed.

[atarione]

3ware needs to develop new chips for their SATA raid controllers they are using ATA bridges on these SATA controllers and of course the controller chips are several years old now.

3ware needs to step it up w/ SATA controllers.

Drivers?

[sjbe]

While I've admittedly not read the entire article (it's really long) I couldn't find much info about drivers. It seems the author basically assumed one would be running windows, which for servers (the most likely place for a RAID array) is a pretty poor assumption. I've tried a number of SATA RAID cards on my linux server (SuSE 9.1) and keep getting driven back to SCSI due to crappy/non-existant driver issues. Thank god for Addonics SATA-SCSI adaptors which work great and have saved me a bunch of money.

It's a nice article comparing performance but without a serious analysis of drivers along with it for Windows AND linux (and Mac if applicable) the article isn't complete. I don't really care which one is fastest if I can't run it on my system.

Re:Drivers?

[LWATCDR]

3-ware has very good support for linux

Re:Drivers?

[ajrs]

read the rest of the article. The fine article says which ones have drivers for different versions of MS, prebuilt drivers for linux, Bsd (which must not be dead), Mac, and if the source code is available.

Re:Drivers?

[carcajou]

Just got in a new IBM XSeries 306 with integrated ServeRaid and dual SATA Drives...there were no drivers for SuSE 9.1...Suse 9.2 found the drives on its own, but at the time of install there were no ServeRaid drivers...ended up mirroring with SuSE...

Re:Drivers?

[rsalvo1975]

Since it's 32 pages long, I won't tell you to RTFA. I didn't read all of it either, but here's a relevant snippet I found while skimming the article.

Besides having a great feature set, the Areca ARC-1120 and its family members enjoy excellent driver support. For Microsoft operating systems, Areca supports Windows 2000 and higher, including drivers for the upcoming x64-versions of Windows XP and Windows Server 2003. Driver packages are available for the Linux based Mandrake, Red Hat, Red Hat Advanced Server, Fedore core, SuSe and SuSE Linux Enterprise Server operating systems, in addition to the source code for each of these. FreeBSD versions 4.2, 4.4 and 5.2 are supported, and driver source code is available for those as well.

Re:Drivers?

[FemmeT]

You're absolutely right about the importance of driver quality. Running all the benchmarks took a huge amount of time (more than four months of testing, 20 hours a day). I just had to limit the scope of the article for practical reasons.

There is some information about driver support listed in the feature comparisons on page 13 and 15.

Re:Drivers?

[AKnightCowboy]

I second the 3Ware recommendation. I've got 3 3Ware controllers at home and 9 of them at work and all of them work flawlessly with vanilla Linux kernels. The volume just shows up as a big SCSI disk. Plus you can move a bunch of disks from one controller to another and not lose your volume. They are seriously the best ATA RAID controllers I've found... pricey, but you get what you pay for.

Re:Drivers?

[Anonymous Coward]

3-ware has very good support for linux

I'm running under FreeBSD 5.3 over here and I'm very happy as well. They may not be stellar performers, but for a small server they're just fine.

Re:Drivers?

[budgenator]

your post in Militants Planned Attack On Indian Software Firms (#11875966) Look, I know exactly two things about Indians: They stole our tech jobs and they constantly try to screw over hard working gambling-addicted Americans by building casinos on their reservations
was the most clueless, stupid and racist thing I've ever had the privalage of modding flamebait -1 in many years!

Re:Drivers?

[MindStalker]

I know I am being WAY off topic. But I think he was trying to be funny. As its Native American's whom we often call Indians that have the casinos. While its citizens of the country of India that are receiving outsourcing jobs. So the poster was pretending to equate the two.

Re:Drivers?

[LWATCDR]

Dud what are you? Some kind of Slashdot stalker?

Re:Drivers?

[budgenator]

I didn't think of it that way at the time, but it might not be an inacurate discription in a more literal manner. Of course I don't have any intention of it being a continuing activity, his post just pushed several of my hot-buttons and just modding him down in an anonymous manner was unsatisfing. I'm wondering what will happen in meta-moderation so far nothing

Re:Drivers?

[AKnightCowboy]

You really do need to get out more. That sound you heard before moderating was the joke swooshing over the top of your head. People like you need to lose your moderation privileges permanently because you do not read the FAQ. If you would have you would see that you're supposed to moderate things up and are discouraged from moderating things down. When I moderate I never moderate anything down unless it's a blatant troll (goatse.cx, penis bird guy, GNAA, etc.). People that use the moderation system and their limited grasp of humor to make a political statement piss me the fuck off. Slashdot used to actually be fun before they made people register for accounts you know. You probably don't remember those days when you could post funny inline images and witty remarks. No, you're just a hand-wringing worry-wart aren't you?

Re:32 pages? No thanks.

[biglig2]

I think you're missing a bit of sarcasm. Robbedoeske writes "Which of the contestants delivers the best performance, offers the best value for money and has the best featureset?"

Well, the answer is, to quote Spaced: "skip to the end".

Areca ARC-1120 is the best card, and the HighPoint RocketRAID 1820A is the best value for money.

Easy!

Adaptec 21610SA

[canofbutter]

Interesting that they didn't review the Adaptec 21610SA. I would have liked to see how it compared to some of the other cards on their list (especially the Areca 1160 (which is the only other 16-port card)). I own one of the 21610SAs and think it's complete garbage (arrays must be less than 2TB, drives configured to operate as independant volumes can not be moved off the card, failure alarm can hardly be heard when using a mere 2 Vantec Tornado system fans (80mm variety). The Areca 1120 they reviewed sounds impressive, and the 1160 sounds like it is equally nice. Too bad they don't point out where to buy these; my Adaptec could stand replacement and someone selling the Area 1160 seems to be hard to find...

Re:Adaptec 21610SA

[Methlin]

Oh I don't know, maybe try the company that makes them? Areca US [where to buy]. A whole 2 clicks from google.
/smartass

Re:Adaptec 21610SA

[canofbutter]

Too bad the site doesn't provide any pricing, warranty information, availability, etc. Looks like one would have to call or e-mail them for such information (which is a little annoying). It's likely because "if you have to ask you can't afford it" (though that's not true when figuring equipment prices when writing grants). /me heads for a phone, complains that the number is long-distance

Re:Interesting that the 3ware offerings performed.

[bill_mcgonigle]

They've been synonymous with quality in the RAID industry for many years.

I've had my share of 3ware cards drop a raid pack and need to be rebuilt from the BIOS, doing nothing special at all but running a RAID-0 as a big storage mountpoint. When the online rebuild tools fail you have massive downtime.

Related Question

[aventius]

I currently have a 4-port sata card for Raid5 but I need more space and was thinking of replacing it with an 8-port card. My concern is the data on the drives. Will swapping the raid cards mess with the data or will the new card recognize the raid5 array and continue to function correctly without any data loss?

Re:Related Question

[beavis88]

From what I've heard, you should assume that you won't be able to swap cards out. You might be able to if you stick with the same manufacturer and they're using the same chipsets, but again, don't count on it.

Re:Related Question

[aventius]

Ok... does anybody have a recommendation. This isn't an enterprise solution... its for my home server. I don't have a tape drive to backup my data and I really don't want to burn 100 DVDs to backup my 500GB of data.

What I want to know is... Lets say I swap the cards and it doesn't work, can I revert to the old card without any data loss?

I'm currently using an Intel SRCS14L. Does anybody have any recommendations?

Re:Related Question

[LurkerXXX]

So you are saying you don't have any backups for your data, and you are relying on only one hard drive going bad at a time to keep all your data safe? Ever heard of a power surge from a broken power supply, or lightening strike? Back up your data. If your data isn't that important, then just plug in the new controller and give it a shot.

Re:Related Question

[carcajou]

I have not had good luck with rebuilding arrays from drives, even though some RAID config utils have this feature...if it was my project, I would not count on being able to save the data on the drives...if you do get the array rebuilt from the drives I would be interested in knowing what 8 port card you used, and the nuts and bolts of how you got it to work...good luck!

Re:Related Question

[aventius]

Let me address your questions/comments orderly:

1) This is a server I setup in my house for my personal webpage, mp3s, movies, and backup for my laptop and desktop.
2) I'm a student ... so beer money is more important than a tape drive... hell one tape is like 4 kegs of beer
3) Relying on only one hard drive going bad at a time? Yes... I have never had a drive fail, and with my current setup of only 4 drives.... no I'm not worried about more than one drive failing at one time.
4) My data is important... and I know how to do this data migration correctly given the correct tools in an business situation but this isn't a business situation.
5) I would like to backup my data but I don't have the money for a tape drive to back up 500gb, nor the time to use DVDs for the backup...
6) I have a UPS for power surge protection but you raise a concern about my power supplies. Can power supplies surge while they naturally burn out and die.... not from external power surges... but from the Power supply just up and dying?
7) I think I'll just setup a second server ( I have a spare motherboard laying around) with the new 8 port card and copy all the data to 4 new drives. Then I'll format the four old drives and add them to the new array.
8) As for backup, any suggestions? I use RAID 5.... should I consider using RAID 6? Or should I just save up and buy a tape drive? (Remember this is a home system... I don't mind have a server go down for a week).

Re:Related Question

[LurkerXXX]

1,2) Tapes are of course way beyond what is called for in your situation. Tapes are for company servers where the data changes regularly and the system needs to be backed up quite often. Most of the data on your drive probably rarely changes, or changes slowly.

3) It can happen, believe me.

4) Home data can still be valuable and worth protecting.

5) Most of your data probably doesn't change that often, so DVDs would be fine. 100 DVDs can be had for around $50. That's not cheap for a student, but is probably worth trying to come up with when you think of how many hours you might spend trying to recreate that data. So:

1) Toss a DVD in to burn.

2) Go to class.

3) Go back to step one.

It won't be long until you have the bulk of your data backed up. Start with the most precious stuff.

6) Yes, a failing power supply can in some circumstances send a power surge through your system. It doesn't happen that often, but when it does, it can wreck a lot of your equipment. Plus a UPS will stop most but not *all* surges.

7) A good idea.

8) Use either DVDs or extra hard drives that are either offline or in another machine. Don't waste your money on tapes for your situation.

Re:Related Question

[Tassach]

I really don't want to burn 100 DVDs to backup my 500GB of data

Get two 250GB external USB drives. This is a very cost effective backup strategy.

External enclosures can be had for less than $30 and 250GB drives are under $140 each. Is your data worth $340?

Re:Related Question

[Skater]

Thanks for your insights, even though I'm not the guy who asked. I'm looking into a RAID setup myself for a video-on-demand system at home using MythTV. (I want to rip my DVDs of television shows to the hard drive so I can watch them without having to flip discs.) Backups of that seem kind of funny - I have a DL DVD drive, but it's going to take almost as many discs as the original DVDs if I leave them in the MPEG format. On the other hand, it takes a while to rip one DVD into the library, so the time savings alone probably makes burning a lot of data DVDs with just MPEG files on them worthwhile. (Plus, then I'd have DVDs I could easily grab to take with me on trips.)

So my real question is, do I want software or hardware RAID? I'm thinking of using a generic SATA card and Linux RAID 5, to keep it cheap and not rely on one hardware vendor. I was thinking of buying 3 250GB hard drives to get started and adding more later...

Re:Related Question

[LurkerXXX]

Real hardware RAID is usually not cheap. Watch out and read a lot of reviews. Oftentimes I've seen cards advertised as hardware RAID, but it was really software RAID being done.

Frankly I don't see the need for the expense of hardware RAID for PVR type stuff. It was your main computer system and you were trying to wring every ounce of performance out if it, it might be worth it to you, but for PVR? I'd go with software.

Re:Related Question

[Skater]

Yeah, and I should've mentioned that this is my server, not my desktop PC. Speed isn't critical, and it's not heavily loaded by any stretch even in it's current configuration, a Duron 850 with 512 megs of RAM. (It'll likely have an Athlon XP 1900+ when I'm ready to set up the RAID drives. The Duron will become the MythTV frontend.)

Re:Related Question

[Yeechang Lee]

I'm looking into a RAID setup myself for a video-on-demand system at home using MythTV. (I want to rip my DVDs of television shows to the hard drive so I can watch them without having to flip discs.)

Having recently built a 2.8TB RAID 5 array for home use, I wrote a long Usenet post on this very topic.

Backups of that seem kind of funny - I have a DL DVD drive, but it's going to take almost as many discs as the original DVDs if I leave them in the MPEG format.

As I wrote elsewhere in the resulting thread, I'm not going to worry about backups for my array; it's simply too expensive and impractical.

Re:Interesting that the 3ware offerings performed.

[PDXNerd]

RAID 0 is not the most reliable thing in the world. Couple that with the unreliability of SATA (yes, I work in a Validation Lab and we go through dozens of these a day) you would *NEVER NEVER* want SATA in RAID 0 storing anything valuable. Swap, sure, but never data! That said, we test dozens of SATA raid controllers as well. The best performer in my experience has been the 3Ware 9500-8. Does it have many advanced features? How many people who will be using SATA raid really NEED those advanced features?

Re:Interesting that the 3ware offerings performed.

[arivanov]

Well... as someone who has both of reviewed 3ware adapters in production I am not amazed. They are nice, but nothing to shout about. They also have LOADS of PROBLEMS not mentioned in the article.

• 8506 SATA series prior to a certain board revision are extremely susceptible to bus noise. As a result you have to find a way to bastardize the PCI bus down to 33MHz and provide additional grounding. Even so, they are likely to cause random system deaths and serious memory corruption in most Opteron MSI and Assus motherboards as well as some other designs. Using in 1U and 2U chassis with riser cards is a no-no for the same reason (exemption for some buffered risers). As a side note, most resellers will try to stuff you with an old board despite the fact that they know about this problem.
• 9506 board and linux driver at least as of 2.6.9 defaults to no read cache, only write cache which is outright daft. It is also the major reason for low performance at least under Linux.

Both are nice cards, but I would not recommend them to anyone who does not have extensive PC hardware knowledge. They are fussy, carpicious and very hard to troubleshoot when they go wrong.

If it doesn't say

[cmefford]

3ware, It's a waste of /my/ time. So I really don't care what the tweakers think. The 3ware cards are reliable, easy to deal with, have brilliant drivers, good software, and they WORK! Always! I have a 5 of them. I have a friend who has 40, he agrees. I use a 2 channel as a backup-to-sata drive, (cheaper than tape), another 2 channel in a IIS server for payroll stuff, 1 4 channel for a mail server and 2 8 channels for file/web. I love'em. Nuff said.

Why would RAID require drivers?

[EnronHaliburton2004]

I'm having trouble accessing the article now, but I couldn't find any mention of hardware vs software-based RAID devices. Could anyone clarify a bit?

Why would RAID require drivers after everything is configured?

It makes sense that drivers would be used for a software configuration tool, because you are doing something which requires special access.

However, it seems that for day-to-day use, a good hardware-RAID device shouldn't require any special drivers anymore. A hardware-RAID card will handle all of the RAID functionality in a way that is transparant to the OS. For example, you have 2 RAID1 disks, the OS should only see a single disk disk. All of the RAID stuff happens behind the scenes.

If the RAID array requires drivers or if I need a special Kernel driver, doesn't that mean that the device is at least partially software-based, which then means that some processes may actually be handled by the CPU, thereby reducing performance.

My experience with RAID devices is obviously a little thin. Maybe I'm misunderstanding, but this question has come up in some LUGs and a few of my workplaces. Just asking for some clarification.

Re:Why would RAID require drivers?

[Wesley Felter]

Even a true 100% hardware RAID controller has to have some way to get requests from the OS to the controller, thus a driver. In theory the RAID controller could emulate something like AHCI that you already have a driver for, but for whatever reason they just don't.

Re:Why would RAID require drivers?

[Anonymous Coward]

All hardware requires drivers, even hardware raid. I don't understand why you think a driver made for a certain card is a "special" driver and makes it software raid.

On top of basic drivers for I/O, hardware raid requires some kind of management tool. That has to talk to the card via a kernel driver. You don't expect it to be so transparant that you can't tell when there is an error?

Re:Why would RAID require drivers?

[carcajou]

This does confuse a lot of people...that's okay. With software mirroring, the OS will spend system resources writing to both disks in a mirror, then spend more resources making sure the disks stay synced. With the hardware RAID, the RAID Adapter driver talks to the OS, so that the OS reads and writes from the adapter. The adapter does the multi-drive write and sync, so the OS does not have to use resources for this...with the added benefit that, in some environments, since the OS sees one drive (actually the adapter), if one drive in the mirror goes down the OS never knows...no reconfig to boot from the other drive, etc.

Re:Why would RAID require drivers?

[Anonymous Coward]

If the RAID array requires drivers or if I need a special Kernel driver, doesn't that mean that the device is at least partially software-based, which then means that some processes may actually be handled by the CPU, thereby reducing performance.

You're confusing 2 things. These adapters are raid cards, and they are disk controllers.

You need a driver to talk to a disk controller, even a plain regular one. Linux and BSD drivers are hard to find for some disk controllers.

You are thinking of some scsi raid enclosures where the enclosure manages the raid, and as far as the computer is concerned, the computer sees one scsi disk, without knowing any details behind the scenes. The computer only needs a regular scsi card. This raid enclosure can be easily used with many operating systems, since the operating system only needs to know how to talk to a scsi disk, and most do.

Re:Why would RAID require drivers?

[EnronHaliburton2004]

Thank you! That explains what I am seeing with some RAID devices.

You answered my question, even though I couldn't phrase my question correctly.

I need to read up on RAID some more ...

Re:Why would RAID require drivers?

[cheekyboy]

I think cards/hw should have built in port 80 tcp web admin on em, no need for error prone driver IO ;)

translate it from a i2c IO or something, I mean if a cheap $30 router can have web admin, so can a $200 RAID controller.

That said, in the absence of a driver, the raid should still 'work' by letting the OS use a default IDE driver.

But I understand the issues of reducing costs by moving some logic into drivers and leaving hardware less complex and error prone. Its just a balance.

My thoughts

[tonsofpcs]

Areca ARC-1120 looks better on each and every page except for the sequential read/write tests where it tends to come in third [I'm just reading off the graphs].
The RAIDCore BC4852 seems fastest for sequential reads/writes.

BOTH of these have linux support. The Areca supports: Mandrake (9.0),Red Hat (7.3, 8.0, 9.0, AS 3.0), Fedora Core (2, 2 AMD64), SuSE (7.3, 9.1 Pro, 9.0 SLES, 9.0 SLES AMD64)
The RAIDCore: Red Hat (9.0, AS 3.0), Fedora Core (1)
The Areca also supports Windows XP and Server 2003 64-bit versions and BSDs: 4.2R, 4.4R, 5.2.1 (incl. source).

Also, the Areca ARC-1160 (they finished testing after the original article was written, so it didn't make it into most of the text) appears at the top of all of the Index/performance tests, except for "Fileserver - Large Filesize - RAID 1/10" and "My SQL - Data Drive - RAID 1/10".

SCSI vs SATA

[sjbe]

SCSI, in its current form, is just opening itself up to becoming antiquated.

Perhaps, though personally I've had far more trouble getting SATA (and IDE) drives to work than SCSI drives and I've used both extensively. Driver issues mostly. SCSI's performance is better in multi-user systems, it's easy to set up, drivers tend to be less problematic especially on systems other than Windows, and it can have more devices attached. People claim it's more reliable though I have no evidence of this, and frankly am a bit dubious of the claim. SATA is also easy to set up and is a lot cheaper, though the drivers are still less ubiquitous than with SCSI and performance doesn't match SCSI yet for multi-user systems. (on a single user system it doesn't matter much)

That said, the next generation of SCSI is Serial Attached SCSI which is compatible with SATA. A SAS controller will be able to use SATA drives if you don't need the extra features of SAS. SCSI isn't going away, it's just adapting.

Re:SCSI vs SATA

[gmezero]

Here's how I seperate SCSI vs SATA. I use SATA RAID setups for video workstations that need large drive space and cheap drives... and I don't care if the drive pops after a year of abuse.

I put SCSI in my servers (RAID or otherwise) when I want the box to run for years and years under heavy load and not have to worry about replacing drives regularly.

With SCSI, your paying for the quality control/quality assurance more than anything else.

From what I understand a good SATA drive has the same TTL quality as a good IDE drive, just faster performance.

Re:SCSI vs SATA

[superpulpsicle]

More like SATA/IDE is the ONLY option for high capacity.

SCSI drives available to the general consumer don't go higher than 180GB. SATA and IDE are way beyond that mark. If someone has a link to a 400GB single SCSI drive, let me know.

Re:SCSI vs SATA

[GigsVT]

the next generation of SCSI is Serial Attached SCSI

Who cares?

Why would you want to make yet another serial interconnect standard, when SATA is perfectly fine?

Oh, I forgot, it's so the hard disk manufacturers and interface card manufacturers can charge 5 times more for a slightly faster and slightly more reliable product... i.e. marketing bullshit.

There's absolutely no reason we couldn't have had 15k RPM ATA and SATA drives by now, other than the obvious collusion and lack of competition among the few remaining hard disk makers.

Re:SCSI vs SATA

[tkg]

Well here's a link to a 300GB drive from Seagate. A bit pricey though at around $1400US street price. Haven't seen any larger drives for SCSI.

Re:SCSI vs SATA

[Anonymous Coward]

Drive makers take their time with hi-cap SCSI drives because they have a reputation for reliability to maintain. They can't just release their latest density drives with a SCSI interface if they expect to be able to charge a premium for the reliability factor. If something fails, your SCSI customers (aka big spenders) will definitely chew your ear off about it, and will likely switch vendors immediately for all future projects. Corporate environs are fickle that way.

ATA and SATA have always been cheap because the home market does not have such strict reliability concerns, and corporate ATA customers are expected to know what they're getting when they avoid paying the SCSI premium. Thus the switch down to 1 year warranties for many ATA drive manufacturers, which was concurrent with all the crazy capacities and experiments you're now seeing in the home market (like 400 GB drives and 16 MB disk caches or whatever other gimmicks they're going after for the hardcore PC 1337d00dz these days). Oh, how I long for the days when all Macs shipped with SCSI ports and you could daisy-chain devices for _years_ before Apple popularized USB (and FORCED its adoption) with the iMac. Prior to that, SCSI was your good-old high-speed Mac interface technology, and you got a nice SCSI drive to boot (pun intended).

Anyway I'm not saying that ATA, SATA, or the SCSI standards are bad technologies. Far from it, I think they all serve a good purpose (but I wish SATA had supplanted ATA completely by a few years ago). I am just pointing out that people rightfully expect more from SCSI because the advantages exist and they are paying for them. The drive manufacturers will charge more for SCSI because they take more responsibility in the production of SCSI products, since its advantages are attractive to customers with huge reliability and performance requirements. SCSI is the perennial high-end, and always will be for as long as it is faster and has the superior featureset (which it always has, ignoring price).

Re:SCSI vs SATA

[Anonymous Coward]

Oh, spit out the sour grapes already. We both know how it is (I don't think I'll be pointing out things you weren't trying to say yourself), but you're ignoring the logic behind the situation just because you want to have your cake (the fastest drives on a slower but newer whiz-bang consumer interface) and eat it too (cheap cheap cheap).

High end customers don't buy ATA because the featureset is inferior to that of SCSI in many tangible ways (devices/channel, interface speed, error checking, etc.). Drive makers use SCSI's advantages to sell their own offerings (high-rotation drives, superior QA, longer warranties). They do this while striking a balance between cutting-edge and proven reliable technologies, based mostly on cost of R&D and production.

So if ATA has these deficiencies that high end customers don't want (and that drive manufacturers can't charge for), why should the drive manufacturers attach their high end products to this tech? That would just serve to cannibalize their server products. No inter-corporate collusion is required here to see that it just makes better sense to keep SCSI prices high for as long as it is the better tech. All the manufacturers do this independently, just because it is the smart thing to do, not to make Joe Leet angry.

Instead, they market less expensive enhancements to low end and home users, like higher capacities and bigger onboard disk cache. After all, how important is latency to the average home user, really? How common is it for the typical home user to require the same low latencies and high transfer rates that a production webserver requires, and how much money do you want the drive makers to lose just to serve those few users in those few situations? After all, guess what, the ones who need faster than ATA/SATA can provide.....buy SCSI.

Re:SCSI vs SATA

[apdt]

Hmmmmmm... One of the major differences between SCSI and IDE is that when accessing an IDE drive, your system processor does a fair amount of the work. With SCSI, this work is offloaded to the SCSI controller and the drive, leavinf the system processor free to do other stuff. This is why SCSI is a much better option for servers (and also why it's more expensive).

I think part of the reason that SCSI has been such a pain is that there have been so many SCSI 'standards'. IIRC, SCSI3 defined 5 properties that the system should have, but mandated that a SCSI3 device had to implement any 2 of these. This is insane IMHO.

Things seem to have returned to some sort of sanity with the current Ultra160 and Ultra320 standards.

This article is a good description of all the SCSI standards.

Re:SCSI vs SATA

[GigsVT]

They do this while striking a balance between cutting-edge and proven reliable technologies, based mostly on cost of R&D and production.

Looks like I've hooked a live one here, a real person from a hard disk company marketing department! Hint: Real people don't talk like that.

High end customers don't buy ATA because the featureset is inferior to that of SCSI in many tangible ways

If it is inferior, there is no technical reason for it to be that way, only a business/marketing reason.

why should the drive manufacturers attach their high end products to this tech?

I think the success of the WD Raptor shows that customers DO want high end drives with standard interfaces.

That would just serve to cannibalize their server products

That's just business-speak for "we want to be able to charge some customers 5 times more for basically the same product"

After all, how important is latency to the average home user, really

Why don't you ask seagate, which has implemented NCQ on even their low end drives?

The writing is on the wall, people are realizing that you don't need SCSI to build enterprise class RAID systems. The hard disk manufacturers that embrace this trend (WD Rapton, Seagate/Maxtor NCQ) will succeed, the ones that continue to fight it will die.

There's absolutely no reason for there to be two different standards other than price descrimination and market segmentation that does not serve the customer.

Re:SCSI vs SATA

[orlinius]

If that's not a coincidence - I just finished a benchmark comparison of SATA vs. SCSI in 2 Dell server configurations to be used uses as mail servers.

I am quite satisfied by the SATA performance but I had trouble with the linux drivers.
I'm very surprised that this study didn't even mention the Adaptec SATA Raid controllers, which are also used by Dell in some of their cheaper configurations.

Here's my benchmark:

http://www.damsys.com/benchmarks/dell-sata-scsi/

I found that in terms of performance the 2 controllers were very close.

Depending on your needs, any of these configurations might be suitable. If you need to have more capacity for the same price you should definitely opt for the SATA Raid setup. If you do not need that much capacity and you want to be absolutely sure in the stability and reliability of your system, go for the SCSI Raid setup, which is based on mature technology with no surprises.

Re:SCSI vs SATA

[GigsVT]

One of the major differences between SCSI and IDE is that when accessing an IDE drive, your system processor does a fair amount of the work. With SCSI, this work is offloaded to the SCSI controller and the drive, leavinf the system processor free to do other stuff.

You need to update your information. This may have been true in 1993 or something.

As it stands today, you are much better off with a northbrige based ATA/SATA controller than a PCI based SCSI card. With SCSI, you are wasting tons of your PCI bandwidth on something that could be coming straight out of your northbridge.

The only advantage to SCSI currently is latency, and that is quickily becoming irrelevant too with the introduction of NCQ SATA drives and controllers, along with some 10k rpm SATA drives.

My experience with 3ware

[dfn5]

I can say from my own experience with 3ware is that it sucks. We decided that we wanted to use S-ATA because we could get a lot of disk cheap. The problem was that these escalade cards didn't do parallel IO very well and by that I mean if one user is doing a long write operation the entire RAID array would go unresponsive to other users. For example if I created a large 20G oracle datafile the entire system would seem unresponsive until the operation completed. I wouldn't even be able to ssh into the server. And this was RedHat AS in case anyone wanted to know.

Moral of this story? You get what you pay for. SCSI should be used for servers.

To be fair, however, I was never able to determine if it was a result of using S-ATA, 3Ware or the linux device driver.

Re:My experience with 3ware

[hackstraw]

I have no experience with 3ware, but I've heard that they are not that great, and that they are basically software RAID that depends on the drivers to do the work for you. This is probably why the system is unresponsive during a large write.

OTOH, I have an Apple Xserve RAID that uses SATA drives with a fibre channel interface. In using it, I cannot tell its not a SCSI system.

Re:My experience with 3ware

[DuBois]

I've had some pretty bad experiences with the 3Ware 8506-4 controllers on a SuperMicro MB running FC2. When running 4 WD2500JD drives as RAID-5, one of them pooped out and the RAID was running degraded. Got a replacement WD2500JD, which happened to have some hard errors on it out-of-the-box. Had to use the mhdd.com utility to zero out and remap the 14 hard errors, then had to reload FC2 before the machine would boot. At least it was just the OS. The data was on a separate partition, and was OK.

Another server running the same 8506-4 card with only three WD2500JD drives in RAID-5 under FC2 had horrible problems anytime a lot of writing was done. It's an email server, so lots of little files were constantly being written. Had to get another drive and rsync everything over to a new RAID-10 configuration (with a loaner computer). The RAID-10 seems to be working fine, with no write slowdowns.

OTOH, our Apple XServe RAID (with an Apple FC card running under FC2 on a SuperMicro MB) is phenomenally reliable, speedy, and easy to administrate.

BTW, The Apple XServe RAID uses standard ATA drives, not SATA. They're Hitachi drives, which may be part of the reason they're more reliable.

WD makes a "RAID" version of the WD2500, the SD, I believe. Probably means that the WD2500JD shouldn't be used in RAID at all. Sigh.

Re:My experience with 3ware

[Henk Poley]

Ehm, wouldn't this be related to the I/O elevator you use? The deadline elevator is known to choke on multiple transfers. You could try CFQ (Completely Fair Queueing), though I don't know how well that works on RAID other than RAID0 and RAID1.

Re:Interesting that the 3ware offerings performed.

[archen]

Just as an aside and sticking up for 3ware. 3ware is one of the few companies that has good driver support for Linux and FreeBSD. As far as 2port SATA mirroring I always recommend 3ware as my first choice - performance is good enough.

Obviously if you're looking at a raid 5 solution, you're moving more towards higher end stuff, so it would be hard to recommend anything that performs poorly there. Rather dissapointing, but probably not that surprising since their SATA cards seem very similar to the ATA cards, so I'm sure they're throwing performance out the window there somewhere =/

Re:32 pages? No thanks.

[Daniel Boisvert]

Good summary. As long as you care about one set of benchmarks and don't give a damn about reliability, real-world performance, or whether the manufacturer will answer your phone calls when the magic smoke gets out of the adapter, go with those cards.

It's not like you'd set up redundant disks for reasons other than short-term performance, right?

(not trying to flame, merely to point out that there might be other things you'd consider before "recommending" a RAID adapter)

OS ???

[Anonymous Coward]

Why on earth dosen't the article say what OS they are testing on?

Re:Comparison of Nine SATA RAID 5 Adapters

[FreeLinux]

Troll my fanny. Have a read of the Linux SATA RAID faq. Notice all the claims that the RAID controllers aren't really hardware RAID controllers?

These claims are like the old claims that winmodems weren't really modems. Remember that? Strangely, now that someone figured it out, almost all winmodems work with Linux now and they are real modems again.

The fact that these RAID controllers are working RAID controllers for other operating systems but, don't work on Linux means that they really are hardware RAID controllers, contrary to the faq's claims. It means that LINUX does NOT support them. It means that no Linux developers have figured out how to make them work, yet.

I'm also willing to bet that Linux inability to work with many SATA RAID controllers is one of the primary, undisclosed, reasons that the majority of the controllers in the article weren't scored.

Oops, there's a problem. Let's just pretend it doesn't exist.

Re:Interesting that the 3ware offerings performed.

[Anonymous Coward]

The Areca claims to have FreeBSD support, does anyone know how it works under 5.3? It offers some interesting features, but every 3ware Card I've tried works without hassle in FreeBSD 5.3.

waste of time and money.

[Anonymous Coward]

I am running RAID 5 in my computer right now.

Linux software RAID. Makes all this crap obsolete except for some specific cases.

I can have as many drives as I want, I can have hot swapability, I can have hot spares and all sorts of fun stuff.

Add LVM on top of that and you have a solution that is much superior then going out and buying any raid controller, except for the most fastest.

Linux software raid is actually VERY nice, I don't know of any OS that has better setup.

Re:waste of time and money.

[Anonymous Coward]

Software raid is slow as hell, especially software raid 5. This stuff isn't designed for your computer, it's meant for servers.

Re:waste of time and money.

[Anonymous Coward]

Linux software RAID. Makes all this crap obsolete except for some specific cases.

Are you out of your fucking mind? You've obviously never worked with a real RAID. And no, eight ten gig seagates inside your homebrew pentium along with the neon lights in the case do not count.

Add LVM on top of that and you have a solution that is much superior then going out and buying any raid controller, except for the most fastest.

I think this says it all. Get off the computer and go back to your room.

Re:waste of time and money.

[ltwally]

"Linux software RAID. Makes all this crap obsolete except for some specific cases."

So, you're saying that somehow your software RAID is calculating XOR bits and such without putting a serious hurt on your CPU and memory? Interesting.

"...I can have hot swapability..."

You're also saying that your motherboard has hot-swap capabilities built into it? Because it takes nothing short of specialized hardware controllers and BIOS's to be able to hot-plug a drive in. (ATA/SATA drive initialization is done during POST. Sure, Linux may do its own thing afterwards, but you cannot escape the legacy limitations of PC hardware.)

No offense... but seems to me that you're talking outta your arse, here. Software RAID offers only one thing in comparison to a good hardware RAID solution: price. In every other aspect hardware wins.. assuming your hardware RAID card isn't a POS.

Re:waste of time and money.

[Phil Wherry]

So, you're saying that somehow your software RAID is calculating XOR bits and such without putting a serious hurt on your CPU and memory? Interesting.

I'm not the original poster to whom you're responding, but the answer to your question is actually kind of surprising (it was to me, in any case).

At any given point in time, your system is in one of three states:

• partially idle (there's unused CPU and disk I/O capacity),
• CPU-bound (the CPU is fully utilized but there's disk I/O bandwidth available), or
• I/O bound (the CPU has spare cycles, but the disk can't provide data fast enough to put them to use).

I suppose that a heavily overloaded system could be both CPU and I/O bound, but it would require a mix of CPU- and disk-intensive processes that isn't usually seen in practice.

Let's ignore the partially idle case, in which there's ample disk and CPU to go around, as it doesn't really matter in this scenario whether the CPU or disk controller perform the XOR operations.

In the case of a CPU-bound process, you're going to incur the additional CPU overhead of the XOR operation. XOR is almost absurdly fast, particularly if the data is in the CPU's cache. I'm pretty sure that modern CPUs execute XOR on at least one byte per clock cycle. But let's say, for the sake of argument, that it takes three cycles per byte. On a CPU clocked at 3 GHz, you'd be able to perform XORs on one gigabyte of data per second if you ignore memory and cache issues. Given moderate memory bandwidth, you're also able to transfer over a gigabyte of data to or from the CPU per second. Given a more reasonable amount of data (say, one megabyte, to transfer), you'd be looking at a CPU impact of around one millisecond to perform the XOR. That's a 0.1% impact at most in a CPU-bound environment, and that's presuming you're doing a megabyte of disk I/O per second.

Now let's look at the I/O-bound case. Here, the CPU is sitting around waiting for the disk I/O to finish up. In this case, it clearly doesn't matter who's doing the XOR operations, since the CPU isn't fully utilized. PCI bus utilization is going to be increased by up to 100% (in the worst-case scenario involving drive mirroring; the worst-case RAID5 scenario is a 50% increase). A typical server's 66 MHz 64-bit PCI bus has a capacity of around 533 megabytes per second (PCI Express increases this dramatically, but let's stick with pessimistic examples for now). At the moment, a SCSI bus tops out at 320 megabytes per second, and those transfer rates are only achievable with at least four drives on the channel and an almost exclusively sequential I/O mix (the best-case numbers for a 15,000-RPM drive are about 100 megabytes/second). So there's generally bus bandwidth to spare.

You raise a number of other points in your note that are potentially issues (hot swappability, for example). But I've become convinced that the CPU/machine performance argument against software RAID really only made sense when CPUs/memory/bus bandwidth were much more constrained.

Re:waste of time and money.

[Anonymous Coward]

You're confused. This is crackpot.org.

Re:waste of time and money.

[Pretor]

Just wanted to assert what you said so I wrote a short test program. You have to compile it without optimizations since then its going to unroll the loop and only do one calculation.

Runnning this on my 1.6 GHz Pentium M takes 31 seconds. That's 129 million instructions per second. Way lower than what you suggests.

Oh btw: I would guess that all the variables fit in the cache of my CPU:)

#include <stdio.h>
#include <time.h>

int main()
{
time_t start = time(0);

unsigned int i;
int i2;
for (i = 0; i < 4000000000u; ++i) {
i2 = 52 ^ i;
}

time_t end = time(0);

printf("Time used %i.\n", (int)(end - start));

return 0;
}

Re:waste of time and money.

[Phil Wherry]

Interesting test - thanks for trying it!

My suspicion is that the CPU is a whole lot faster at performing an XOR on a relatively large block of memory than it is at doing a correspondingly larger number of XOR iterations on a few bytes. In particular, the MMX instruction set extensions were designed to handle logical operators on large blocks efficiently. By doing it a little at a time, you spend a correspondingly large amount of time handling your loop counter and associated comparisons.

Re:waste of time and money.

[etymxris]

If I'm doing software RAID-0, does this do XOR? If not, what other penalties might a software RAID-0 have over a hardware implementation? I'm actually curious as I plan to set up a software RAID-0 myself within a week or two.

Re:waste of time and money.

[Anonymous Coward]

When you load the md drivers in Linux they will show the speed of which they can do the RAID parity checks.:

raid5: automatically using best checksumming function: pIII_sse
pIII_sse : 1016.000 MB/sec

This is on a Pentium 3 running at 500MHz.

Re:waste of time and money.

[drsmithy]

So, you're saying that somehow your software RAID is calculating XOR bits and such without putting a serious hurt on your CPU and memory?

Almost certainly. For example, on my rather humble (by today's standards) 350Mhz Pentium 2 server:

raid5: measuring checksumming speed
8regs : 644.400 MB/sec
32regs : 334.800 MB/sec
pII_mmx : 783.200 MB/sec
p5_mmx : 817.600 MB/sec
raid5: using function: p5_mmx (817.600 MB/sec)

I don't have a machine handy to check, but more modern CPUs should have checksumming speeds well and truly into the gigabytes per second. That's faster than any disk subsystem you're going to shoehorn into a PC.

This is assuming you're using RAID5 - which is dismally slow on disk writes for reasons much more significant than the mythical "XOR CPU overhead". Overhead for RAID1 should be utterly insignificant - software *or* hardware.

The *CPU overhead* of software RAID on any remotely modern system is barely even measurable. Anyone arguing against software RAID (and, in particular, RAID5) by using the "CPU overhead" argument, is about seven years behind the times (and in the case of RAID5, demonstrating a deep misunderstanding of why it is slow).

Sure, Linux may do its own thing afterwards, but you cannot escape the legacy limitations of PC hardware.)

Hot plug is a raw hardware (and SATA supports it, even if all controllers don't) and OS+driver issue. The PC bios and "legacy limitations" aren't an issue since no current OS accesses the hardware in that way.

Software RAID offers only one thing in comparison to a good hardware RAID solution: price. In every other aspect hardware wins.. assuming your hardware RAID card isn't a POS.

No, hardware RAID offers only one major advantage in comparison to software RAID and that is transparency. To the OS, a hardware RAID array is just another block device, so things like disk failures will have no impact on things like the booting procedure (which with software-only RAID can cause problems).

A good rule of thumb: if you want RAID for your system drive, use hardware RAID (for transparency) and if you want RAID for your data, use software RAID (for better performance, better flexibility and better reliability).

Re:waste of time and money.

[Anonymous Coward]

People who write software-raid-drivers as suboptimal as what an unoptimizing c-compiler makes of your code still have a lot to learn. Measuring "C-instructions per second" is almost completely meaningles.

Re:waste of time and money.

[MarkRose]

You raise some excellent points. However, what about when a drive fails and you need to rebuild the array? Do you want your bus saturated for hours while your CPU busily performs an XOR operation for each byte on each drive? I'd much rather have a dedicated chip on the RAID device to handle this, although the software RAID is an option if you keep your drives on a separate bus (some motherboards support such).

Re:waste of time and money.

[ErikZ]

The man just said 1GB per second to saturate.

Which is far beyond the writing speed of the fresh hard drive you've put in. 1/10 that if you're lucky.

So, 10% CPU to rebuild a HD, at 100MB per second 400GB hard drive would take a little over an hour to rebuild.

Re:waste of time and money.

[ErikZ]

You forgot to add

"And you can put all that money that you were going to buy the RAID card with, and buy more/bigger HDs"

Re:waste of time and money.

[jepaton]

You have to compile it without optimizations since then its going to unroll the loop

This unfortunately means you end up benchmarking loop overhead.

You could either enable loop unrolling ONLY, or keep optimisations disabled and unroll by hand.

On my 1.4 Mhz Athlon, your code runs in 17 secs (no opts) and in 0 secs (level 3 opts) :P

Unrolling by hand, 20x, gives 5 secs (no opts).

Re:waste of time and money.

[cowbutt]

# cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 15
model : 2
model name : Intel(R) Pentium(R) 4 CPU 2.40GHz
stepping : 4
cpu MHz : 2412.523
cache size : 512 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm
bogomips : 4784.12
# modprobe raid5
# grep 'MB/sec' /var/log/messages
Mar 10 14:12:18 xxx kernel: pIII_sse : 3092.000 MB/sec
Mar 10 14:12:18 xxx kernel: raid5: using function: pIII_sse (3092.000 MB/sec)

Lots of hardware RAID controllers use the Intel i960 as the embedded controller. I've read that 66MHz models are only slightly faster than a PII-400.

TRUE raid?

[fire-eyes]

All I care about is if these are 100% raid, unlike a seemingly increasing number of cards. In windows you might do alright, but anything else, look out.

In linux you will be treating such cards as a software raid array. Kind of defeats the point of buying "hardware" in the first place.

Wankers (the manufacturers).

Re:TRUE raid?

[Anonymous Coward]

A card that is as you put it "Not True Raid" are BIOS backed. They are just as slow on windows as on linux. Usually you would use LVM to make raid with such cards, as LVM is highly optimized unlike the vendor drivers.

Re:TRUE raid?

[stu42j]

All I care about is if these are 100% raid, unlike a seemingly increasing number of cards.

Yes they are.

Re:TRUE raid?

[stu42j]

Ok, I lied. Three of them are not hardware RAID.

Do You Speak Printer-Friendly?

[alexburke]

http://print.tweakers.net/?reviews/557

Onboard sata controllers are usually software

[Mustang Matt]

I have an onboard sata raid controller on two of my 1u servers. So I configured it and it's bios said, hey you've got a mirrored drive using these two physical disks.

Linux sees then as completely seperate hard drives. Turns out the sata raid controller relies on a windows driver and is nothing more than software raid. I'm not even sure it's accellerated in any way.

So I just used linux software mirroring and it works fine. (Had to use a sarge nightly to recognize all the hardware.)

Re:Onboard sata controllers are usually software

[afidel]

EVERY single adapter in the review was a true hardware RAID array controller. They all have at least four ports and can all do RAID5. Please read the fine article before spouting off about unrelated cheap schleck.

It had to be said

[Anonymous Coward]

Can the establishment counter the attack of the newcomers?

2000 Quatloos on the newcomers.

Re:Comparison of Nine SATA RAID 5 Adapters

[stu42j]

Except that all the controllers reviewed in this article are true hardware RAID with good linux support. In fact, the article even spends a good deal of time discussing the issue.

If you want RAID on Linux (SATA or otherwise) you use true hardware RAID if you can afford it or Linux software RAID if you can't. Why bother messing with that fake-RAID crap, even when Linux is supported?

Re:Comparison of Nine SATA RAID 5 Adapters

[Anonymous Coward]

You don't know what you are talking about. Most cheap raidcontrollers are software! The vendor drivers just hides the disks and presents one device to the OS. But all the raid calculations is done in the driver. On on the card. The cards reviewed here do have a on borad CPU and thus are true raid controllers. And Most of them. (Not promise and Raidcore) provide source drivers for linux that does use the onboard processor.

Re:32 pages? No thanks.

[MindStalker]

Yummy, looking at that makes me want to make a much needed replacement for my existing server. Can anyone suggest a good AMD 64 rackmount case/MB combo that fits nicely with this, the ones I can find are designed with hotplug scsi in mind.

Beware hardware RAID

[puke76]

Sure everyone buys a few spare drives.. but make sure you buy more than one RAID card. If the RAID card goes, unless you replace it with an identical make and model, you can kiss your data goodbye.

That's what I like about software RAID on Linux - you can mount the array on another linux box if you need to.

Have yet to see a good comparison between low-end hardware RAID and Linux software RAID..

Re:Beware hardware RAID

[jdew]

This is not totally true. Ignoring the SATA stuff as I've no experience with it first hand.

The Compaq SmartArray series uses the *SAME*FORMAT* for it's arrays in it's 6302 that it did on the original smartarray.

I have personally moved a raid array from a SmartArray 2P to a 2DH to a 4200 to a 5302. And it's all functioning perfectly.

Re:Beware hardware RAID

[tialaramex]

Anecdotal because I'm not paid to do this stuff..

We bought a 3ware controller for a large and somewhat valuable datastore (high resolution images of Alan Turing's personal papers which include all the text available elsewhere plus handwritten annotations, scribbled diagrams, etc.)

In the end I only used it as a fast and not particularly full-featured ATA controller, running Linux software RAID on top because it was not only _faster_ in every test I could think of, but also simpler to set up and maintain.

There aren't many published comparisons of this sort of thing precisely because the low-end (PCI card etc.) hardware RAID manufacturers lose, and that means they won't be renewing their advertisements in your publication any time soon. In _theory_ they should win if your application is both disk I/O and computationally heavy, and thus you can't spare any CPU to run software RAID, but I've never seen a benchmark that could demonstrate this apparently obvious result with real hardware.

Re:Beware hardware RAID

[justins]

If the RAID card goes, unless you replace it with an identical make and model, you can kiss your data goodbye.

If you are dumb enough to use RAID as a substitute for backing up, that is.

Re:Beware hardware RAID

[owlstead]

Well, that might save your data, but the availability is still a problem. You can have a great backup (probably using tapes for a RAID array) but if you need to get an exact copy of a RAID card on sunday you are in trouble.

Anyway, it's rather hard to backup something like 1 TB of data using cheap storage solutions which is why RAID as backup is currently viable. Just don't use it for sensitive information and beware of software issues...

Re:Beware hardware RAID

[justins]

Anyway, it's rather hard to backup something like 1 TB of data using cheap storage solutions

It's not really, anymore - or at least, it shouldn't be.

For what I used to pay for a single DLT tape a few years ago I can get an external USB drive of greater size. The problem, and it is a big one, is that most backup software doesn't have any idea how to properly take advantage of something as advanced as hot-swappable, external, fast USB mass-storage.

Re:Beware hardware RAID

[Skuld-Chan]

I've actually swapped 3ware controllers of similar (not entirely exact) make and model and it has always been able to rebuild the raid and just work. Same with my ICP (scsi) raid controllers.

My biggest problem with software raid is that it is very alpha software. Problems occur, and there's little documentation for the new tools when you need help - and trust me - when you're data is on the line you won't want to putz around with something that isn't well documented if you don't have a support network to help out.

Re:Beware hardware RAID

[CAIMLAS]

When was the last time you looked at software raid?

I've found software raid to be better documented and have more features than the various IDE RAID cards out there, bar none. Granted, I've not seen all the cards, but it seems better in my experience.

Re:Beware hardware RAID

[Anonymous Coward]

most backup software doesn't have any idea how to properly take advantage of something as advanced as hot-swappable, external, fast USB mass-storage.

Go look at veritas backup exec 10.

Re:Beware hardware RAID

[ErikZ]

How about cheap enough?

Oh, and lazy.

Look, I just want a bigass drive that can hold all my MP3s and any video I record with my video card. I don't want to mess around with tapes.

If the machine explodes violently, well, that's life. If a drive dies, I get it replaced and don't lose any data.

Re:Beware hardware RAID

[Rich0]

The few years ago part is relevant here.

A 300GB (compressed) SDLT tape costs about $50 these days.

You aren't going to get a 150GB drive for anywhere near that cost.

For 1TB of space, those tapes would run you about $500 (uncompressed space). Equivalent cost for internal hard drives would be about $900. For external drives you'll pay more.

Tapes also work better in libraries, which make management easier.

I guess it depends on the situation. Hard drives are probably faster, but more expensive. On the other hand, tapes require an expensive drive. If you were just making one backup per year the drives would probably be cheaper just from a hardware cost perspective. On the other hand, if you ran two backups per day and retained them for a month, and retained one backup per week for a year and one per month for a decade, the tapes are going to work out much better most likely.

Re:Beware hardware RAID

[justins]

A 300GB (compressed) SDLT tape costs about $50 these days.

You aren't going to get a 150GB drive for anywhere near that cost.

I see them as cheap as that all the time on slickdeals and some of the other sites. Usually a price that low involves a rebate, but still.

Re:Beware hardware RAID

[Skuld-Chan]

Last time I looked at it was a year ago when all I did was reboot a file server and I could never mount that raid file system again :(. 500+ gigs gone, and no-one seemed to be able to help.

Re:Comparison of Nine SATA RAID 5 Adapters

[Anonymous Coward]

The fact that Linux now supports jackasses that buy software modems does not make them real modems. They still leech on the CPU.

Re:Comparison of Nine SATA RAID 5 Adapters

[Anonymous Coward]

Well, given that the winmodem hardware provides neither mo or dem functionality, it's not really a modem, is it? It's a DSP coprocessor or something. Software modems may well be a good thing, but without the right cooperation from the CPU, a winmodem isn't a modem.

Hardware RAID is something that I plug into a bus, that presents a number of connected disks as one big disk to my hardware. If it requires a special driver for my operating system in order to perform some of the RAID functions, it's not hardware RAID.

The combination of one of these cards and its windows driver is a mixed hardware/software RAID, but if the card does not present itself as something that I send SCSI or ATA commands to in the same way as I would address a "normal" disk, it's not a hardware RAID card.

Re:32 pages? No thanks.

[Fred_A]

Well, I read somewhere that the blue ones are usually faster.

Maybe in Dilbert.

Re:Comparison of Nine SATA RAID 5 Adapters

[swv3752]

Winmodems are not really modems. Have you compard the performance of a softmodem and a hardmodem?

In anything less than ideal conditions, which means pretty much anywhere where broadband is not a viable option, softmodems suck. You'll be lucky to manage a 28.8k connection. Hook up a hardware modem and suddenly you get a 53k connection.

But as for the software raid, what do you thinks happens in Windows? There is a driver that fools the config utils in Windows into thinking the array is one drive. What happens in Linux? There is a driver that allows one to setup a software raid and configure an array. So both Windows and Linux do the same thing but Linux is more honest about it.

Re:32 pages? No thanks.

[__aazofn1209]

I can say from my recent experience setting up a SATA-RAID storage server that the needs and preferences of people using this technology are so varied that it would be hard to trust anyone's "executive summary".

Thirty-two pages is a lot, but it's also a big investment of time, money, and possible future frustration to setup and manage a storage array. More than likely, anyone who is really interested in this review is actually interested in some of the nitty-gritty content.

What I'd like to know is why so many SATA RAID controller reviews happily include solutions that are not "true hardware RAID" (i.e. 90% of the non-3ware controllers) but mostly exclude the high-quality many-port SATA controllers that are best suited to software-only RAID setups.

For example, the Promise SX8 8-channel PCI-X SATA controller has good linux support, but it's hard to find so much as a whisper about it when people talk about storage arrays.

Adaptec?

[Anonymous Coward]

Why didn't they bench anything from Adaptec?

my 2 cents

[Ankou]

We have used 3 LSI 150-6 MegaRaid Cards and I must say that its the most increadible card / bang for buck you can get. Works perfectly in linux (Slackware 10.0 - 10.1 in our case), uses either the megaraid or megaraid2 (for those that want verbose information) right from the stock kernel compile. In each server we put in 6 Seagate SATA drives 250 GB each, totalling an impressive 1.2 TB total space. For under a grand (card + 6x 250 GB drives) you cant get a cheaper more reliable alternative. The thing aint slow either, consitantly get access of 100 mbps transfer speeds or more (hdparm tests / benchmarks). Initialization is almost instantanious, and while its doing the background inits (after the initial quick init), you can already access the entire contents of the full raided container. Do yourself a favor and grab one of these cards, you'll wonder why you stuck with the almost 3x price of scsi. Newegg.com has em for I think 290 bucks for the 150-6. Pay the extra money for the 150-6 its worth it. Optional battery packs available as well for the card.

The Article Never Explains What RAID 5 Is

[windowpain]

Now you could argue that a car review in Car and Driver doesn't bother explaining what a transmission does but RAID is several orders for magnitude more complex and esoteric.

There are so many different flavors of RAID it can be hard to keep them straight if you're not working with them every day.

Anyway there are good explanations of RAID here and here.

Re:The Article Never Explains What RAID 5 Is

[justins]

Now you could argue that a car review in Car and Driver doesn't bother explaining what a transmission does but RAID is several orders for magnitude more complex and esoteric.

Are you kidding?

RAID 5 can be explained in a few pages - the math, the implementation, the whole bit. Have you ever seen a technical drawing of a transmission? Modern slushboxes are about the most advanced mechanical engineering application that the average person ever comes in contact with (when they aren't at the airport).

You won't find an article that does most of the issues involved in designing and implementing a transmission justice. I know you just meant it as an example, but still. :)

Re:The Article Never Explains What RAID 5 Is

[Anonymous Coward]

Now you could argue that a car review in Car and Driver doesn't bother explaining what a transmission does but RAID is several orders for magnitude more complex and esoteric.

You've obviously never tried to rebuild a transmission. There are so many different flavors of transmissions it can be hard to keep them straight if you're not working with them every day.

Anyway, there is a good explanation of transmissions here.

Re:Comparison of Nine SATA RAID 5 Adapters

[operagost]

Well, winmodems really aren't modems without the software to support them. The same goes for "software" (really host-based) RAID controllers. These controllers perform their calculations on the host CPU, so writing a driver for these is a much greater task than writing one for a "hardware" RAID controller (one with its own CPU).

Re:32 pages? No thanks.

[Tassach]

Tyan makes some very nice AMD-64 rackmount barebones systems.

Re:Comparison of Nine SATA RAID 5 Adapters

[killmenow]

It means that LINUX does NOT support them.

How you got modded Insightful is beyond me. Are you a troll? Did you RTFA? I didn't read it all either. But I read enough to catch this much: LINUX does support ALL of the cards in this review. Two of them (the Promise and RaidCore cards) come only in binary packages but the rest have Linux source available.

I, personally, would completely avoid any card manufactured by Promise or Highpoint as I've had crap luck with them in the past. They're just not very good cards, imho. And I'm not talking about their performance in Linux. I'm talking their performance in general. They're crap by my estimation regardless of platform. After losing data on my Windows 2000 box becuase of a crappy Highpoint card, I'll never buy another.

Anyway, your assertions are not even germaine. You point to the problem with "trick-BIOS" software RAID cards, which have been around for years and are not exclusive to SATA-RAID. They are shit cards, period...have been from the day they were made. Most of the cards in this review, however, are true hardware based SATA-RAID cards.

And, again, they all are supported on Linux. 3Ware, for example, has been a bastion of Linux support for ages.

As for the whole winmodem issue, who cares? What has it to do with a freaking troll blathering incorrectly about Linux not supporting SATA-RAID cards? Besides, the fact is, winmodems are NOT real modems. They're telecom interfaces, but not modems. You need software to make them modems. And I'm not talking about driver software to give access to the cards' functions. I'm talking software that has to implement the modem functionality itself...because the modem functionality doesn't exist on the "winmodem"...because it's not really a modem. Just because we now have linmodems.org and such to provide that software, it doesn't automagically make them "real" modems.

I did read the article.

[Mustang Matt]

I was specifically addressing the parent posters mention of Dell including SATA in their cheapest server line. Chances are those are on board software controllers and not hardware controllers as seen in the review.

So take a chill pill, a deep breath, count to 10 and smile.

Re:I did read the article.

[sumdumass]

I just ran into this today with a promise raid card. It apears that there are actualy 3 kinds of adapters and they follow thru into both SATA and PATA (ide). there is first a totaly hardware implementation wich is real expensive but does not need any system software to function properly. Next there is a firmware implementation wich is hardware excelorated but requires a system driver to function properly. And finaly there is a pure software raid wich uses standard operating system and disk devices to creat the raid.

The firmware version is the one dell ships and the one you had described in the erlier post. This confused the hell out of me when trying to get a promise fastrac 100tx to work in linux today. There is no 2.6 kernel driver yet and i had to settle with setting a software raid and just using the add in card as a way to conect more drives. This server is only a network storage server so i guess it shouldn't matter all that much.

It was confusing and misleading though when i bought the cards because it did say hardware raid controler and it did say it was compatible with linux. Just not the version i was running and it is firmware controled (operating system driver) as oposed to a actual hardware controlled were it is totaly transparent to the operating system.

I dunno exactly why i'm babling so much about this except for maybe i just wanted to share my experience and let you know it can get real confusing really fast with terminoligy that sounds interchangable but have deeper meanings. Nothing was misrepresented with the promise card i bought rather i didn't understand the statments made or the extent of the meaning of the statments. It still allows me to do what i want, i just have to tackle it differently.

Re:Interesting that the 3ware offerings performed.

[fragzillax]

Gotta agree on the 8506's being flaky. I could not get them to work reliably with 3 different mobo's (one was a supermicro, very nice server board), after calls to supermicro, 3ware, and much hair pulling, I purchased the Highpoint listed in the article, and have been very happy with it, especially considering the price delta for the 8506-8's... So I currently use the HPT, and have 2 8506's sitting in boxes that I'll never use because of all the trouble they caused. Maybe time to head out to the 'ol gun club...

Re:Interesting that the 3ware offerings performed.

[RFC959]

I am going to chime in with my damnation of 3ware's cards too. We have about 20 8500s and 8506s, with either Maxtor or Seagate drives. The things are horribly unreliable. Almost every day at least one array needs to be rebuilt. On a few occasions, we've even seen the controllers spontaneously lose an entire array - just poof, not accessible anymore and not visible through the administrative tools. Reboot and there's the array again.

Most of our support has been through a VAR (who sucks too, but that's a separate rant), but when we talked to 3ware and told them how we were using the arrays (for database storage), they immediately went, "Ooh...uhh...that's...not really a good idea..." Even they admitted that the SATA arrays are really for very light-duty use only. (I blame our old VP of technology, who always wanted to go the cheap route on everything.)

Oh, and if you want to upgrade the controller's firmware? 3ware tells you to boot off a DOS-formatted floppy. This is not enterprise-level stuff.

32 pages? Linux ?

[GodWasAnAlien]

32 pages of ?? and nothing regarding compatiblity with Linux.

Actually, I only searched the conclusion....

Re:32 pages? Linux ?

[stu42j]

All of them support at least RedHat. Some of them are binary only. This is listed on the feature charts:

http://www.tweakers.net/reviews/557/13
http://www.tweakers.net/reviews/557/14

Wow... RAID-6 cards!

[doormat]

Finally, a controller that supports RAID 6. RAID 6 is just like RAID 5 but with an extra parity drive, so that you can have two drives (instead of just one) fail in an array and be OK. RAID-50 is slightly less robust (two drives on the same RAID5 chain can break and then you're up shit creek), but faster (for the same card implementation).

The interesting thing is that the Areca card is infact SATA-II. Things like NCQ, and port multipliers can really elevate usefulness. Buy a cheap 4-port multiplier card and

Re:Wow... RAID-6 cards!

[Palshife]

Nitpicky, I know, but it's not an additional parity drive, it's an additional parity calculation that's distributed among the disks evenly, just as in RAID 5. It's sometimes called two dimensional parity.

http://www.acnc.com/04_01_06.html

Re:Wow... RAID-6 cards!

[doormat]

heh, yea, my post originally had "adds a parity drive in another dimension" but I figured that too many people wouldnt get it.

SATA Reliability

[Anonymous Coward]

SATA Drive Reliability is 1/2 that of SCSI or FIBRE... Don't believe the hype... It makes a nice backup to tape alternative, but should never be used for a production environment.

If you go with SATA you'll be spending time and money replacing components and restoring from backup.

Re:SATA Reliability

[Anonymous Coward]

1) Prove SATA drive reliability is half as good as SCSI drive reliability.

2) Even if you have the proof, so what? Properly configured RAID and a watchful admin pretty much make single-drive failures a moot point. This is, after all, one of the main advantages of RAID (excepting JBOD/RAID0): if a drive fails, you simply replace it, with no down time and no data loss.

Below a certain MTBF, drive mech reliability is not a huge issue. The main thing holding back IDE/SATA RAID hs been the availability of enterprise features (such as online expansion or battery-backed RAM) in non-SCSI RAID cards. The tweakers.net roundup shows that this is ceasing to be an issue.

Re:SATA Reliability

[bratmobile]

Would you care to provide some facts? Or should we just take your word on this?

Nearly all drive manufacturers sell the same physical drives with lots of different interfaces. SATA *drives* are no more or less reliable than SCSI *drives*.

Re:Interesting that the 3ware offerings performed.

[chthon]

I had the rebuild problems also with my Promise SX6000 controller, but strangely I had it more in the beginning, and when I used ReiserFS on my RAID-5 system (running Red Hat 9).

The same goes for upgrading the controllers firmware.

It seems that both companies make the most money of a market based around small companies who want to give themselves the air of professionality by implementing their own storage solutions.

couldn't agree more

[FreeUser]

I have to agree. I've deployed 3-ware SATA controllers at work and at home. Disk recovery has been straightforward, driver support for Linux excellent (it's in the stock kernel sources), and performance more than satisfactory.

It would be nice if one could expand the array "hot," rather than having to copy data around and redefine/reformat the array, but in terms of reliability in protecting my data against disk faults (I've had several disks die, and replacement was a breeze, with zero downtime). As others have pointed out, for most purposes, Linux software RAID is both faster and more robust for recovering from catastrophic hardware failures: put the disks in a new computer and big up lvm2. For those applications that need hardware RAID, I've found the 3-ware escalade cards (both SATA and the older EIDE devices) to be excellent.

Re:Interesting that the 3ware offerings performed.

[Cramer]

Yeah, I thought that was really f'ed up too... no read cache. what the f*** is the point of having so much memory on the card then? And let me tell you, 128M+ write-back cache is very bad idea.

So, where can you buy the Areca cards?

[bourne]

The review looks nice, I'm convinced. If I want to buy an Areca card in the US, where would I go?

Google doesn't help. Pricewatch doesn't help. Tom's Hardware didn't provide an answer that I could see. Nothing on eBay but palm trees. What appears to be the US distributor has a "Where to buy" link that points to the Taiwanese site which points to... the US distributor.

Re:So, where can you buy the Areca cards?

[Anonymous Coward]

Try looking for TekRam cards instead. I think they are the redistributor for areca cards. Same numbering. ie: Tekram 1120 rather than Areca 1120. Anyone know for sure if they are the same? I know the tekram uses the areca chips, but are there actual areca cards that are not the same as the tekram's?

Informative, but badly written

[Gondola]

While informative, my brief read-through of the article revealed numerous typos or spelling mistakes. The pictures and research here is cheapened by the lack of proper editing and proofreading.

Samples from TFA:

"...caused by differences in I/O processor and I/O controller performance, cache memory, available bus bandwidth etcetera."

If you're not going to use the traditional abbreviation "etc." at least use it properly; "et cetera."

"You can't make judgements by simply..."

Judgment is spelled with one "e".

"Thanks to the onboard network connection, the management software can be directly available over the netwerk withough the help of server-side software."

"Network" and "without" are both misspelled, in the same sentence.

"It can be considered quite and achievement that..."

"And" instead of "an". A common typo, easily rectified with proofreading.

These typos/misspellings are all on the last page of the review. I could go back through the review and find more, but I think this sampling is enough to testify to the poor standards applied to this article.

Re:Informative, but badly written

[a-freeman]

It was translated from an article originally written in Dutch. Pretty sure english was not the translator's first language

Re:Informative, but badly written

[bratmobile]

So, someone made a few tiny benign mistakes when *translating* an article. Grow the fuck up and just ignore them.

Re:Informative, but badly written

[Gondola]

Hey, you can use four letter words in a semi public venue and not get spanked by daddy. Hooray for you!

When an article is referenced on Slashdot, that article is up for critique and discussion. That is what I did. Or are you part of the Slashdot staff now and giving orders about what kind of discussion may or may not take place in response to article postings?

Dumbass.

Re:Informative, but badly written

[Dutch Hakker]

OMG I have never seen such a nitpicker in my life. You know what. Next time you write something do it in Ducth then we can all laugh at your stupidity. They translated 32 pages to english, while they normally write in ducth, so the rest of the world could profit from what they published. Bashing them as so called badly written review when they come here doesn't really encourage them to do more reviews in english. It's not our native language so cut us a little slack, besides your english isn't that great either since "judgement" is correct and "network" is not a name thus it is allowed to be written in lower case. So go back to school moron and learn your own language first.

Re:Informative, but badly written

[Gondola]

Your English is much worse than the article. You get an F.

"Judgement" may be in online dictionaries, but it is a variant of "judgment" and is (or at least used to be when I received my Bachelor's in English from the University of Michigan a few years ago) *wrong*. Ask any English teacher.

I never said "network" should be capitalized. I said it was spelled incorrectly with two e's (ie, "netwerk") see my post instead of making things up, thanks.

If you want to translate something poorly, don't post it on Slashdot unless you are prepared for criticism. Sort of the same way you don't run for President of the United States if you're George Dubya unless you want flash movies making fun of your speech fuckups all over the internet.

I applaud those that are fluent in more than one language. It's a worthy skill to have. However, in this forum, I am free to critique the articles presented. This site is obviously an English site. Deal with it.

3ware 9500 is dog slow..

[slashmojo]

I have been using it in raid5 on a file server running centos3.3 first and then fedora3 and its just terrible.. to the point of unusable.

There is a thread on redhat bugzilla relating to the 3ware performance problems that is almost a year old and still no resolution afaik (although I haven't yet tried rhel4).

Just about to order a replacement so this article was well timed indeed!

Re:3ware 9500 is dog slow..

[donio]

Try using them in JBOD mode and software RAID-5, that works much better here and gives more flexibility too.

Not hardware RAID

[jgarzik]

Note that several of the cards reviewed in this review are not hardware RAID. SATA RAID is famous for being non-RAID controller + RAID software driver.

See my SATA RAID FAQ for a listing of the most common SATA chipsets which are sold as RAID, but are really software RAID (a.k.a. "fake RAID").

I'm also rather amazed that this wasn't mentioned in the review, but I admit I did not read all the of the 32 pages.

Re:Not hardware RAID

[stu42j]

Only three out of nine are not hardware RAID. And the article does mention which are which and it spends a whole page (page 3) on discussing the differences.

Re:Not hardware RAID

[drsmithy]

See my SATA RAID FAQ for a listing of the most common SATA chipsets which are sold as RAID, but are really software RAID (a.k.a. "fake RAID").

Not that it really matters, as long as you have appropriate drivers.

The protocol, the cards and the drives

[owlstead]

Saying that S-ATA is worse than SCSI is mostly not due to the protocol (see the various S-ATA is not as good as SCSI comments) but due to the cards, and more importantly, to the drives.

SCSI cards and drives have been created specifically for enterprise use. That means that they perform, and that they last. The only way to compare these technologies is to use an expensive S-ATA controller with fast hardware XOR, large cache and controllers and drives that support command queueing. Furthermore the drives should turn rapidly, have average seek-times around 5ms and very high MTBF figures. The current crop of S-ATA cards are nowhere in this league. There are the converted Raptor drives from Western Digital Raptor series, and that's it.

This does not mean that S-ATA RAID configurations are useless:
- redundancy
- storage of large data sets (read mp3, mpeg4, porn and wares)
- raid 0 for video (large sequential data)
The cards reviewed have the benefits of tools, the use of a RAID as boot windows drive and - of course - the added serial ATA connectors. For high performance web-servers or multi-user setups, look elsewhere.

My RAID fantasy: 1394/USB2 Raid hub

[swb]

I'm sure that me and the few drunks I've managed to hoodwink with this concept are the only market for it, but why not a USB2/1394 hub that's actually a RAID controller?

The hub could present whatever defined logical volumes to the OS as additional mass storage devices on the hub, and a configuation application would be all that was needed since the logical volumes would be presented to the OS as generic mass storage devices.

I think this could have a real market; while the bus would certainly be a limitation in performance (perhaps 1394b would help), it:

* Wouldn't require a massive case with internal bays and power taps for the drives. (S)ATA RAID is cheap, but scaling beyond 3 or 4 drives is a huge challenge in all but the biggest cases. Using external connectors like 1394/USB2 would solve this easily.

* Wouldn't require any drivers beyond existing USB/1394 generic mass storage support. Yes, you would need a special application to configure the hub's logical volumes or to perform stupid RAID tricks, but beyond that you wouldn't.

* Portability to other systems, either in the event of a host failure or, since it doesn't require drivers and once configured, it could be moved to another platform that only supported the generic mass storage device.

* OK, speed would suck, but it's about adding big, reliable mass storage with a trivial interface, not about transfer rates. The hub could actually have distinct USB/1394 channels to individual ports, since it's not really a _real_ hub and the host OS wouldn't see the individual disks, just the defined logical volumes presented as mass storage devices.

I think this would be great for "backup" applications or other small-time/home user data warehousing (keeping your native DV-AVI files, DVD backups, CD backups, MP3 backups, yadda...) Tape is nice, but SDLT or LTO drives are expensive, as are the media. For $600 you can do better than half a terrabyte of RAID-5 disk, but you need almost an entire PC to house internal disks.

Given how cheap RAID cards are, I can't believe that merging RAID into a hub would be all that expensive, especially since you're actually removing a lot of the disk control logic from the controller.

that's great, why don't you...

[Anonymous Coward]

that's great, why don't you do a better job translating the text to, say, dutch?

rewriting 32 pages to a foreign language, and missing only a few grammar and spelling errors - i'd like to see you do better.

"judgment also judgement"

[slashmojo]

http://www.answers.com/judgement

FYI ;)

Re:"judgment also judgement"

[Gondola]

I believe this definition to be a response to public usage. Languages do evolve over time; when I was in school, however, judgEment was *NOT* acceptable.

Whew

[falser]

Just last week I bought the 6ch LSI card, and will be recieving the rest of my drives this week. After reading the relevant parts of the article I was a little worried I made the wrong choice. But your comments have eased my paranoia. Thanks.

Re:Interesting that the 3ware offerings performed.

[bill_mcgonigle]

RAID-0 is unreliable against drive failures. This is a given.

That should never mean that it's unreliable because the controllers flake out!

For storing large amounts of easily-recreatable data on a local machine it's a very acceptable way to keep the costs low.

Areca...

[Epsillon]

Never heard of them until now. Reading TFA and visiting their web site has made them unforgettable. 10/10 for getting it right and 15/10 for supporting most of the major open source OS out there (I nearly said all, and risked getting modded as a troll), including FreeBSD. We could do with more companies like these.

Tekram in the US

[charnov]

it appears these are carried in the US under the Tekram brand...

ARC-1110

Re:Tekram in the US

[canofbutter]

Still no 1160, unfortunately. This is not that big of a deal, though, since it seems that the 1120 (and perhaps the rest of them too) have multi-adapter support (something the Adaptec 21610SA does not)

speaking of linux and raid

[jaxon6]

I came across this project called SCST
http://scst.sourceforge.net/
It lets you take direct storage(lvm, raid, plain disks) or files on the system and serve them out over fibre channel to clients. So you can take 4 sata disks of 200GB each, RAID5 that up and get 600GB usable space. Break that 600GB into ten 60GB partitions and serve that out, and you have absolutely failsafe storage for your systems. Windows systems can use any old supported(most are) fibre channel card($25 on ebay), plug into a linux box that is running SCST(linux box needs a qla2200/qla2300 card, also cheap on ebay) and the Windows host sees a 60GB scsi drive. I'm testing it out right now, but it seems to solve the problem of remote storage cheaply and sanely.

MOD PARENT UP

[anti-NAT]

Because Jeff knows what he is talking about. He can't afford not to :-)

IDE Can do hotswap!

[Anonymous Coward]

You're also saying that your motherboard has hot-swap capabilities built into it?

IDE can do hotswap just fine. Either you can use the hotswap tool in Linux (by Tim Stadelmann) or you can use Windows 2000 and later.

In Windows you simply stop the device before removing it, then you choose scan for "hardware changes" and it will re-detect new devices on the IDE bus.

IDE does support disconnected use and bus reset. A bus reset will re-detect the devices on it. Nothing magical with it, nor do you need specialized hardware.

What you should use is a harddisk rack to about electrical problems when you pull or connect the cables.

Re:Comparison of Nine SATA RAID 5 Adapters

[stu42j]

OK, I take that back. 2 of the controllers are completely software based and one is 'hardware assisted'. I guess I just ignored those completely when I read it the first time.

Even still, the article does explain that these three are not true hardware controllers and spends some time explaining what that means.

Re:32 pages? No thanks.

[Anonymous Coward]

Actually, if you just jump to the end, you are reading an ad for Areca. Tweakers did not compare equal products or configurations. They exclaim the performance of the only card fitted with 1GB of cache and test older versions of other products.

Grats to Areca for pulling off a marketing ploy.

If you want real information, read the article completely.

3ware supports 1GB, but Tweakers didn't bother doing that test. Tweakers blames supply?

Re:SCSI vs SATA, geek history

[Anonymous Coward]

Too bad they didn't call it "Serial Attached SCSI Implementation", or SASI. They would have come full circle (SASI, for "Shuggart Associates System Interface" was the first SCSI).

Re:Interesting that the 3ware offerings performed.

[Skuld-Chan]

Both are nice cards, but I would not recommend them to anyone who does not have extensive PC hardware knowledge. They are fussy, carpicious and very hard to troubleshoot when they go wrong.

I wouldn't say that. For one thing I had my 3ware controller running just fine with debian out of the box and I didn't have to tweak anything. While it might not be the best performing it at the very least has excellent support in Linux, for example smartctl can work with 3ware (but not any other raid controller) not to mention all the 3ware management tools run in Linux as well.

Also - 3ware's support group is second to none. They answer the phones right away, have excellent warrenty service and they speak english.

Even though this fly by night review site gave 3ware poor marks they will always get my business because I've always had good experience with their hardware and service/support.

Also - on a side note - 3ware is very very very good about maintaining compatibility with firmware. I've never seen a firmware release break a raid controlled by 3ware.

Re:My RAID fantasy: 1394/USB2 Raid hub

[Ahnteis]

Sort of like this?

Or were you looking for something to do RAID 5?

Re: USB 2 version (RAID 5)

[Ahnteis]

Other message was a simple raid 1 solution.
This one is RAID 5.

I'm sure there are others. :)

Re:Interesting that the 3ware offerings performed.

[ckedge]

We've seen similar things with a pair of LSI-Logic 150-6's that we've got. One is a lot less reliable than the other, but both still drop a drive at random and have to be rebuilt. The most twitchy one is now running a striped set a lot more stable, but it's stupid not being able to run the RAID-5 we bought it for. We've now RMA'd two maxtor HDDs from that array, so possibly they were contributing - but it's frustrating finding a drive dropped, rebuilding, and finding that everything is ok again - and doing that week after week.

And of course LSI won't take it back unless they see some *definitive* problem, and their line is "manufacturer hard drive utilities don't always reveal all transient problems".

What was worse was that it wasn't until Suse 9.1 or 9.2 that LSI had a full suite of drivers and tools for use from within a running instance of Linux - back when we had Suse 9.0 a drive dropped meant that you had to take the system down to the bios to do the rebuild. No utility was available to do it while the system was live! All the while they claim "linux/Suse compatible". Phhssssshhtt.

Good thing we've got the *pair* of systems running as a HA set.

A lot cheaper than the $10-30,000 commercial options, but requires WAYYYY too much babying by the admin. Took forever for the admin to learn all the config and to get the HA set up as well. Not sure if we're ahead or not.

Re:32 pages? No thanks.

[ErikZ]

I'm using a Promise SX4 card, and their linux support has been horrible.

I'd be very surprised to find out that this is completely different with their other cards.

Re: USB 2 version (RAID 5)

[swb]

These must be real new, as the last time I looked (last summer?) I never found any.

I've run across a couple since then (google, doh!) but they're almost always complete enclosures, which I guess isn't bad from a space conservation perspective, however, it does automatically limit you to 3 or 4 disks out of the box -- with the small advantage that the drives are probably on a couple (S)ATA channels internally, and not further slowed by a USB/FW connection.

A hub-only device would be superior, though, as you'd be able to add disks and add/grow the RAID container as needed, vs. fscking around adding a whole new box when you wanted to add capacity.

What's not made clear from the few I've looked at how RAID is actually handled in the box or on the host. To pass my test, the device would have to _only_ present logical volumes to the host as mass storage devices and use actual hardware RAID internal to the box.

What concerns me is that the devices aren't real hardware RAID; they're presented to the host as some nonstandard USB device and you have to install a USB driver that does software RAID to the drives.

Re:My RAID fantasy: 1394/USB2 Raid hub

[cowbutt]

I've used Linux's built-in software RAID to access RAID0 and RAID1 arrays that were built from two USB-attached PATA drives (i.e. transplanted into USB caddies). The bottleneck was the USB 2.0 bus (performance was about half what I got when the same drives were previously attached to an on-motherboard Promise PDC20276 controller), but otherwise worked completely as expected. Further, Intel USB 2.0 controllers were about 10-15MB/s faster than NEC USB 2.0 controllers.

Re:My RAID fantasy: 1394/USB2 Raid hub

[swb]

I don't doubt that the USB2 bus would bottleneck, but software RAID isn't what I'd want (and annoyingly, even XP won't let you do RAID over USB2 disks anyway).

I'd want actual hardware RAID so that RAID sets would be truly portable to anything supporting USB2 mass storage devices. Software RAID is truly unportable without moving it to a system that supports the exact RAID software system. Given the moving-target nature of Linux, I'd call a Linux software RAID completely non-portable.

I'm not saying its inherently bad, it's just not what I'd look for, although for the hub I dream about, an embedded CPU running linux and using Linux RAID would be fine.

Re:My RAID fantasy: 1394/USB2 Raid hub

[cowbutt]

This is, as far as I know, what the larger USB discs (e.g. Lacie) do; a RAID0 array between two (or more) smaller discs.