RAID level for backup server

Portal Home > Knowledgebase > Articles Database > RAID level for backup server

Posted by Mopman, 01-24-2012, 08:35 AM
I'm sure this has been asked a million times, but I did a search and I still can't make my mind up on this - would appreciate some input from those more experienced. I'm setting up some servers that will be storing backups. I wouldn't say performance is the number one consideration - it's backups, so obviously reliability trumps. But, it's going to be doing a lot of them, so dog slow isn't gonna cut it either. Is the added reliability of RAID6 really what it's cracked up to be? I've heard rebuild times&load can be bad enough that the array is basically useless when rebuilding anyway, and surely another drive could potentially take a dive in that time. Should I consider 10?
Posted by Mopman, 01-24-2012, 08:46 AM
Forgot to mention - these aren't going to be particularly fast drives, either.
Posted by rds100, 01-24-2012, 08:47 AM
How much space do you need in the first place? Is this going to be for your own backups, or you want to offer backup services?
Posted by Mopman, 01-24-2012, 08:53 AM
Not just my backups. (Any solution will of course be tested and abused before going live, this is just my initial poking around for advice.) Most of the backup services I see around seem to be using RAID6. Does this mean that the problems with it are overstated, or that those guys are using hefty drives? Seems unlikely given their usual cost per gb, but hey.
Posted by rds100, 01-24-2012, 09:54 AM
When it comes to our own backups i prefer to do RAID1 only. Much easier to recover in case of some (catastrophic) hardware failure - like just pull one disk and use it on it's own, even without a raid card. But when it comes to providing backup service and the price per gigabyte matters i don't know... maybe RAID6 is the answer, or maybe ZFS instead of RAID?
Posted by bqinternet, 01-25-2012, 10:48 AM
We use RAID6 for our backup servers. If you're using a good RAID card, you'll still have usable performance during a rebuild. Drives shouldn't be failing very often anyway. If you do have another drive fail during the rebuild, then you'll be glad that you're using RAID6 instead of RAID5. It would take 3 failures to bring down a RAID6 array. With RAID10, it depends on which specific drives fail, so you can go down if the wrong 2 drives fail.
Posted by euroservers, 01-25-2012, 06:03 PM
I would recommend RAID 10. A reliable option. RAID 10 using a long-time.
Posted by SoftDux, 02-19-2012, 09:35 AM
Use RAIDZ 1 or RAIDz2 if you need more redundancy. It's much more redundant that normal RAID, since it can detect, and correct errors before you get "silent softwre corruption". ZFS also includes deduplication which can save a LOT of space if you have many identical files spread out over the drives. i.e. if you had to backup 2 or 2 identical XEN Virtual Private Servers, then the OS files on the VPS' would only be backed up once, instead of 3 times.
Posted by Softsys Hosting, 02-19-2012, 10:22 AM
RAID 6 is what we use along with 1 Global hotspare - if you have large number of HDD's, RAID 6 makes more sense compared to RAID 5.
Posted by harget, 02-19-2012, 10:24 AM
I'd recommend raid 5 or raid 10. Really depends on how sensitive the data is to you.

Add to Favourites Print this Article