Is raid5/raidZ1 dead for SSD as well?

[Drogo]

Knowing that raid5/raidz is dead due to URE on rebuild/resilver issues, I was wondering if that's also applicable to SSDs? I just figured that different technology would have different failure modes.

Thanks.

edit: I accidentally a word.

 

[cyberjock]

Uhh... depends on the URE of the SSD doesn't it? If your URE compared to the size mathematically works out then RAID5 might be dead there too. You'd have to look up the numbers and do the math for yourself to know if that applies for a given model and brand.

 

[TremorAcePV]

Uhh... depends on the URE of the SSD doesn't it? If your URE compared to the size mathematically works out then RAID5 might be dead there too. You'd have to look up the numbers and do the math for yourself to know if that applies for a given model and brand.

Wouldn't it be fine on SSDs?

The major portion of the problem is that rebuilding a RAID array means quite a lot of reading and writing. SSDs are generally small and from what research I've done, most have a URE of 1/10^16 or great (some are 10^17 and even 10^20), so yeah.

Lower sizes with (orders of magnitude) lower URE rates would probably mean RAID 5 is fine for SSDs. Just something else that's interesting: http://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte

 

[TremorAcePV]

Double Post. The forum has been laggy lately. Not sure if it's me or the server.

 

[DJ9]

I think its the forums. It's been really laggy here also. Click on message to read, wait 8-15 secs and message appears.

 

[Sir.Robin]

After the first disk failure, you are still running an unprotected array with SSD's in raidz1 too until you have it replaced and resilvered.

 

[Drogo]

Yes, but that's not why raidz1/raid5 is "dead" on spinning media. That issue is due to rising URE rates going along with larger drives and longer re-silver times. Otherwise, raidz1/raid5 would have always been a poor choice. I mean, you're always in a degraded state until you fix what went wrong. Even with a hot spare (which freenas doesn't support if I understand correctly), you're unprotected until the resilver process completes.

 

[Sir.Robin]

URE is one reason. Unprotected array after first failure another reason. Long resilver/rebuild time on larger drives yet another reason. Either way, Raidz1/Raid5 suck. :p