On Mon, Mar 8, 2010 at 17:14, David Gibbs <david@xxxxxxxxxxxx> wrote:
I recently saw this on Slashdot ... and since IBM i supports both RAID and SSD's, I thought it might be interesting ...
The comments are a better read than the article itself - it boils down
to a few basics:
Most RAID controllers shipping today can be maxed out by a few SSDs.
IBM's sizing guides already reflect this - for example, a 16 bay drive
bay in a DS8000 should only be populated with 4 SSDs, because of
controller bandwidth issues - and yes, this makes SSD adoption even
more expensive, as those drive bays including controllers don't come
cheap.
An interesting read:
http://www-03.ibm.com/support/techdocs/atsmastr.nsf/5cb5ed706d254a8186256c71006d2e0a/97dfe761d71fee7e862575a7002105ad/$FILE/DS8000%20SSD%20WP%20v1.1.pdf
Most enterprise SSDs use SLC and have double the capacity you need -
this eliminates wear leveling issues, since you can just map new pages
and clean other pages later. Of course, cheaper consumer SSDs like
Intel's X25-M series only have 15% spare capacity and use MLC flash.
100% spare capacity found in enterprise drives also alleviates the
need for TRIM, which means that the OS does not need to be modified to
TRIM empty pages.
What this article is about is build-your-own server setups - you don't
have configuration guides for these, and need to be aware of it. If
you're buying tier 1 servers, you don't need to worry. IBM, HP and
DELL already thought about all these issues.
midrange.com
