Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm setting up a backup storage device- This machine has Windows Server 2008, on a separate boot drive.

It has 8x 1TB drives, and uses a hardware RAID card.

My question is, which RAID configuration should I go for?

Initially, I was going to go with RAID 5 across all 8 drives, however members on serverFault have advised against it. I was just wondering why?

Some people have suggested 2 lots of RAID 5 configuration on 4 of the drives, then striping them...

I want to maximise the storage space, as this is a backup unit - will store SQL backups, Acronis Images, files, etc... It won't be for public access, so the I/O won't be that high I wouldn't think.

share|improve this question
7  
ZFS is what you really want :) –  Bill Weiss Nov 2 '09 at 19:01
add comment

9 Answers

up vote 16 down vote accepted

Fore pure cold backup I'd say RAID5 is fine. RAID5 will take a considerable amount of time to restore on 8x1TB SATA drives, but given the chances of something breaking at low IOPS that's probably not critical. If you want the added bit of safety, go for RAID6 which will give you the ability to survive two simultaneous disk failures.

share|improve this answer
3  
I'm currently using a RAID6 with 8 1TB drives, and it works excellent. –  phuzion Nov 2 '09 at 13:57
6  
+1 for Raid 6 suggestion –  Aaron Nov 2 '09 at 15:02
1  
+1 RAID 6. In this application, performance usually isn't as important. If you're worried about recovery time, depending on your data, RAID 10 with compression may yield the storage you need and shortening the rebuild time. –  duffbeer703 Nov 13 '09 at 1:10
    
another +1 for RAID6. I've got 7x750GB in RAID6, and it works pretty well (and I've had drives die a couple of times, and it recovered fine). –  James Cassell Nov 13 '09 at 2:36
4  
There is about a 60% chance of an unrecoverable read error when rebuilding an 8-disk RAID5 array consisting of 1TB drives. RAID6 and RAID10 are the only safe traditional options in this scenario, but what you really want is RAIDZ2. More about large RAID5 arrays and unrecoverable read errors. –  Skyhawk Jun 2 '11 at 17:40
show 3 more comments

Here is a good article about: Why RAID 5 stops working in 2009

Here's a quick of summary:

  • Disk Failure: With 7 disks you have ~20% chance of seeing a disk failure in one year. It's quite likely you'll see a failure over the operational lifetime of the storage.
  • Read Failures: "SATA drives are commonly specified with an unrecoverable read error rate (URE) of 10^14." This means that for a hypothetical 12TBs of storage you're likely to see read failures. This is just a function of the increasing amount of storage that is commonly being used. More storage, more read failures.

It works like this according to the author: More disks (increased probability of disk failure) + more disk space (increased probability of a read failure) = Increased probability of both events happening at the same time. What will kill your RAID array is when a disk fails and while your array is rebuilding you have a read failure.

The author's solution: RAID-6

Here's a comment that seems to sum it rather well:

The key point that seems to be missed in many of the comments is that when a disk fails in a RAID 5 array and it has to rebuild there is a significant chance of a non-recoverable read error during the rebuild (BER / UER). As there is no longer any redundancy the RAID array cannot rebuild, this is not dependent on whether you are running Windows or Linux, hardware or software RAID 5, it is simple mathematics. An honest RAID controller will log this and generally abort, allowing you to restore undamaged data from backup onto a fresh array.

Editor's Notes: You could just use quality SCSI or SAS drives...

share|improve this answer
1  
Good article, and good point. I run into a lot of folklore and tribal knowledge about RAID that's just not supported by mathematics. It's critical to understand RAID logic and probabilities. –  pboin Nov 2 '09 at 15:55
    
There is a good point raised in this article but I cannot agree with conclusions. Author states that RAID-6 is not a solution either and I wold question validity of this statement. –  dtoubelis Nov 23 '11 at 15:45
add comment

If you opt for RAID 5, you will suffer a heart-stopping panic for a day or so while the array is rebuilt to include its replacement. If you go for RAID 6, when one drive fails you can afford a further failure during the rebuild, so that heart-stopping panic is reduced to mere mild alarm.

If it were I, I'd go for 6 -- but then I have problems with my blood pressure sometimes as it is...

share|improve this answer
add comment

My recommendation is go with RAID-6 without any doubt. Especially considering that you are using 1TB drives.

Here is my experience. During past 4-5 years we were running 3 RAID-5 arrays with 2TB drives - 2 by 5x and 1 by 10x (I hope I'm not to cryptic here :-). During this time we experienced 3 or 4 hard drive failures. Only once array was rebuilt without any issues. All other times there were errors during rebuild. One of them was catastrophic and costed us $12,000 in fees to data restoration company. The severity of a failure highly depended on intelligence of RAID software/firmware. The catastrophic failure was with Linux software RAID. Recoverable failures were with Adaptec hardware controller. (It essentially marks only a stripe as bad if it discovers an URE and then continues. fsck after rebuild and you are back in business. You may loose file or two but not the entire array.)

Since then we migrated our data to RAID-6 arrays with good hardware controllers. We experienced first drive failure a few month ago and recovery was a breese.

Many will justify use of RAID-5 vs RAID-6 by performance reasons (I'm partially quoting my own answer to another thread here), ... this performance difference is THEORETICAL. From my experience the REAL performance difference between RAID-6 and RAID-5 is negligible or non-existent. I'm talking about hardware RAID with decent controller with at least 8 drives, and system with enough memory to provide good caching. In this configuration write speed is likely be limited by SATA/SAS/SCSI bus bandwidth...

Another downside or RAID-6 is cost of this extra drive. But let me put it in perspective. You would probably pay $200-500 more for and extra drive but please compare this against cost of downtime or recovery fees if something goes wrong... Just consider this as an insurance premium :-)

share|improve this answer
add comment

I'd say go for Raid5 over 7 disks, and keep one as HS. But everything really depends on the raid controller you're using, and on the type and quality of your drives.

share|improve this answer
1  
Why would you waste a drive that could easily be used to improve capacity or fault tolerance? –  Nic Nov 19 '11 at 4:21
    
if you think hotspare disks are a waste, there's not much else I can say to you. give it time, you'll see the use for HS if you stick around long enough –  dyasny Nov 22 '11 at 10:54
    
I'm willing to be convinced, so I created a new question on the subject. –  Nic Nov 22 '11 at 20:58
add comment

You'll need to figure out how much your data is worth, in real world dollars, and how annoying or costly a failure will be before you can think about storage strategies. A hint is that the answer can't be "no loss ever," because that has infinite cost.

Use that cost/time/data to guide how long you can tolerate a rebuild, or how many hours the system can be down, or how much money you want to spend on hardware.

share|improve this answer
add comment

On my filer I run a software raid 5 across eight 500 GB s-ata drives, which works great. Read performance near 160 MB/s write performance around 90 MB/s. Recovery time is about 20 hours. Athlon X2 4200+ and nvidia chipset.

As this is to be used for backup storage, I don't see the need for raid 6. It's unlikely that two drives on the backup raid will fail within a few days, at the same time as production data are damaged or lost.

share|improve this answer
add comment

RAID5 accross 8 disks should be fine except it will have poor write performance as it has to calculate the perity bit. Read performance should be really great though.

I am pretty sure the maximum recommened number of disks for RAID5 is 12, but it all depends on how your using the array if you can use that many, the more disks you add the slower your rebuild and writes will get.

share|improve this answer
1  
You are wronjg. Adaptec suppors more than 12 discs on a Raid 5. –  TomTom Dec 17 '10 at 17:55
    
There is a difference between what is supported and what is recommended as practical for a production environment. I was making a recommendation not stating a quantitative fact. –  ITGuy24 Feb 1 '11 at 22:01
    
I've never seen a number that high for "recommended" number of disks tbh, I've always gone by 6 or less just because disks love to die in groups. More than 12 exposes you to a lot of potential data loss. –  Matthew Nov 23 '11 at 16:49
add comment

If it is purely for backup purposes and IO doesn't matter, raid 5 should be fine, raid 6 would be even better but you'd be losing storage space. How critical is your data? You could also do raid 5+0 comprised of two 4-disk Raid 5's. This would let you lose 2 disks, give you better IO, and leaves you with about 6TB usable, but there's a good possibility your controller doesn't support it (usually requires higher end stuff). Raid 6 is slowest of all of the options, lets you lose 2 disks, and will still leave you with about 6TB usable.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.