Boston Linux and Unix InstallFest XXXII Saturday February 28, 2009

[bene-Gk2boCrsRs1AfugRpC6u6w at public.gmane.org (Ben Eisenbraun)]

On Mon, Feb 23, 2009 at 04:28:12PM -0500, Jerry Natowitz wrote:
> A year or two ago Google published results of a big study they did on 
> disk reliability.  I don't remember much about it, except that they 
> found that RAIDs are usually built using disks from the same 
> manufacturing lot, and that failure modes are often quite similar for 
> disks from a particular lot.  This results in a higher probability than 
> expected that a disk failure will involve multiple drives.
> 
> I read this to mean that RAID 6 or RAID 1+0 (sometimes called 10), or 
> possibly 5+0 should be used rather than RAID 5.

Robin Harris from ZDNet/Storage Mojo had a few posts on the Google paper
and RAID 5 in general:

Google's Disk Failure Experience 
http://storagemojo.com/2007/02/19/googles-disk-failure-experience/

Why RAID 5 stops working in 2009
http://blogs.zdnet.com/storage/?p=162

NetApp Weighs In On Disks
http://storagemojo.com/2007/02/26/netapp-weighs-in-on-disks/

Which has this nice incendiary line:
"RAID 5 today verges on professional malpractice"

Fun stuff.

-b

--
irrationality is the square root of all evil.       <douglas hofstadter>