[clug] Is this logic correct? [12Tb RAID-5 array unlikely to rebuild]

[Daniel Pittman]

Adam Thomas <adam.lloyd at gmail.com> writes:
> 2009/3/28 steve jenkin <sjenkin at canb.auug.org.au>:
>> Anyone care to comment if this logic is correct or not?
>> (Large drives in RAID-5 likely to be unable to rebuild)
>> Seems simplistic to me...
>
> I can't really comment on this better than Daniel Rutter (of Dan's
> Data) does on his blog:
>
> http://dansdata.blogsome.com/2008/10/23/death-of-raid-predicted-film-at-11/

Daniel gives a very good summary.

>> NetApp apparently spruik their double-parity RAID scheme as a solution.
>
> Linux's MD system supports a double-parity RAID scheme under the name
> of RAID6. See the mdadm documentation for details.

RAID6 is the standard double-parity mechanism, but be aware that
calculating parity is a little more expensive than RAID5 math in normal
use, as well as the extra write penalty.

That math can be disastrously slow in degraded operation, though, so you
really don't want to run RAID6 where you need high performance even when
the array is degraded and/or rebuilding.

You can also use some variation on RAID10, which gives you vastly better
probabilities since you are now looking at two adjacent drive failures
before you have trouble, not just any two drive failures.


Also, death of RAID predicted, film at 11, is a good summary of this,
since it has been a prediction that sells copy for a good while now...

Regards,
        Daniel

Coincidentally, "death of flash storage predicted" sells a lot of copy
at the moment...