Raid 1+0

[Matthew Seaman]

On 04/19/16 22:43, Lowell Gilbert wrote:
> Correlation is not causation.
> 
> It's not a big stretch to imagine that two nearly identical mechanical
> devices, operated in nearly identical conditions, might wear out in a
> nearly identical way at about the same time. There is no need for one
> drive to affect the other.

Interestingly this argument is quite a bit stronger when applied to
SSD's rather than spinning disk drives.  SSDs we know have a limited
number of refresh cycles for any one memory cell.  As members of a RAID
array, they're going to see very similar patterns of activity over their
lifetimes, so they're actually quite likely to wear out at a similar time.

> A fair number of people believe that this in fact occurs. I've looked
> for evidence on the subject, and I haven't found anything (beyond
> anecdotes) for or against the possibility.

Mechanical drives are a lot more affected by external differences like
vibration or temperature changes, which will act to increase the
variability in their lifetimes.  SSDs are intrinsically /more/ reliable
during their working life, but their lifetime tends to come to a much
more sharply defined end.

	Cheers,

	Matthew (who has been spending far too much time going to the
datacentre to replace drives recently.)




-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20160420/c6646062/attachment.sig>