[CentOS] OT: What's wrong with RAID5

Thu Oct 1 04:20:36 UTC 2009
John R Pierce <pierce at hogranch.com>

Stephen Harris wrote:
> "Almost always" is very dependent on the disks and size of the array.
>
> Let's take a 20TiByte array as an example.
>   
...

he did say 'very large'.

note, raid10 has another parameter...  say you have a 20 drive raid10 of 
1TB drives (10TB total usable).  if one drive fails, a rebuild only 
requires reading one drive and writing the hotspare replacement, this is 
fairly quick compared with the massive restripe operation  of a raid5.

and, if during that rebuild operation, another drive fails, there's only 
a 1 in 19 odds of it being the mirror of the previously failed drive if 
we assume failures are a totally random occurance (yeah, ok, if we 
assume that a drive is more likely to fail when its being accessed, then 
the odds are soemwhat higher tha mirror would fail then another drive in 
the array....  but, an array that does periodic sweeps on idle storage 
will greatly reduce the possibility of this by 'discovering' a failing 
drive much sooner.