Ok so I'm in the process of building a new server to replace my clunky old box and am faced with that age old decision - RAID 1 or 5?
Now everything I'm reading lately is saying RAID5 is a bad idea and should be steered clear of, with the recommendation usually being RAID10 instead. I don't need the performance boost of striping so it'd just be basic RAID 1 mirroring in my case.
Now the logic seems to be that, if a drive fails in a RAID 5 array then, once the drive is replaced, the rebuild process puts a very high load on the remaining drives, which increases the probability of a second drive failing, at which point you're screwed.
Whilst I get that, surely the same applies to a basic mirror too? The entire contents of the surviving drive would still have to be read in order to rebuild the mirror so surely there's just as much chance of failure?
Cost is obviously a big consideration and, given that I need a decent chunk of space, it feels very galling having to halve the effective capacity with a mirror when I could get more bang-for-buck with a parity array.
If there are real, tangible benefits to the mirrored approach then I'd still consider it but I'm struggling to justify it right now.
Any thoughts?
Now everything I'm reading lately is saying RAID5 is a bad idea and should be steered clear of, with the recommendation usually being RAID10 instead. I don't need the performance boost of striping so it'd just be basic RAID 1 mirroring in my case.
Now the logic seems to be that, if a drive fails in a RAID 5 array then, once the drive is replaced, the rebuild process puts a very high load on the remaining drives, which increases the probability of a second drive failing, at which point you're screwed.
Whilst I get that, surely the same applies to a basic mirror too? The entire contents of the surviving drive would still have to be read in order to rebuild the mirror so surely there's just as much chance of failure?
Cost is obviously a big consideration and, given that I need a decent chunk of space, it feels very galling having to halve the effective capacity with a mirror when I could get more bang-for-buck with a parity array.
If there are real, tangible benefits to the mirrored approach then I'd still consider it but I'm struggling to justify it right now.
Any thoughts?