[coyotos-dev] RAID sucks

Eric Northup digitale at digitaleric.net
Mon Jul 23 10:09:43 EDT 2007


Jonathan S. Shapiro wrote:

[...]

> The marketing droids would have you believe that RAID is an answer to
> all of your storage wrongs. Bullshit. The most common cause of drive
> failure in modern enclosures is heat. When multiple drives share an
> enclosure, they tend to fail about the same time. We experienced this on
> Friday night, when two drives died within about 3 hours late one night.
>   
Actually, as I understand the recent large-scale disk drive reliability 
studies (two were presented at a recent FASTA), drive failure was *not* 
strongly correlated with temperature within their operating temperature 
range.  That doesn't exactly match my experience or intuition either, 
but their sample sizes are 4 decimal orders of magnitude larger.  The 
strongest predictor of failure was to have another drive with the same 
manufacturer/model/lot fail (ie, don't buy matching pairs), and that 
does match my experience - and yours, it sounds :-(

I have switched to drive mirroring between internal and external drives.
> The limiting factor in recovery would be funny if it wasn't so
> irritating. Once half the drives are gone, you may as well upgrade all
> of them. Unfortunately, all of our local computer stores are sold out of
> 750G SATA drives. They should arrive on Tuesday, and we'll rebuild
> Tuesday night.
>   
I don't know how much this matters to the overall failure rate, but 
there is a significant loss in how much (and how affordably) data can be 
recovered once you start using disks with >1 platter (such disks also 
produce more heat).  With single-platter drives, data rescue shops can 
swap the platter into an equivalent model's electric/mechanical assembly 
fairly easily and read the data off it.  With multi-platter drives, the 
track alignment tolerances generally mean that once platters are 
removed, they can only be read by expensive, slow magnetic microscopes, 
and sometimes not even then.

-Eric


More information about the coyotos-dev mailing list