On 11/30/10 13:56, Robert Bronsdon wrote:
On Tue, 30 Nov 2010 20:21:06 -0000, Brad Tilley <b...@16systems.com> wrote:
Also, I just noticed that the high-end Intel SSDs claim 2,000,000 hours
MTBF. I wonder why they market that number and then say "3 year
warranty". There's only roughly 26,280 hours in a three year period.
I wonder how many will fail inside the 2,000,000 hours while running
your script then how many will fail sat on a shelf for that time period.
The disk you've (for any given you) purchased will fall somewhere
between these.
While not related to the MTBF discussion, quite a lot of testing has
been done with SSDs by some of the PostgreSQL folks and as a result
there is a Big Warning on the PostgreSQL wiki
http://wiki.postgresql.org/wiki/Reliable_Writes
That links to two separate articles:
"See SSD, XFS, LVM, fsync, write cache, barrier and lost transactions
[1] (Vadim Tkachenko) for an introduction. Their consumer drives such as
the X-25-M G2 have been reported to be even worse [2] (Evan Jones)."
[1]
http://www.mysqlperformanceblog.com/2009/03/02/ssd-xfs-lvm-fsync-write-cache-barrier-and-lost-transactions/
[2] http://evanjones.ca/intel-ssd-durability.html
You can find much, much more in the postgresql mailing list archives.
In short, while there are SSDs appearing on the market that do correctly
report safe writes, the Intels are not among them.
Jeff Ross
