Quote:
Originally Posted by brentleary
Based on recent characterizations of multiple SSD drives using iometer, their is a greater penalty for non-contiguous, random write accesses than with rotating media.
|
Mmmm - that appears to depend on the quality of the equipment - and interpretation of the data.
Have a read of
this - it's long but detailed.
NOOP would have to be a "no brainer" for the IO-scheduler - (I had thought it had a (tunable) delay for consolidating the I/Os, but a quick look at the code would seem not).
But I'd still reckon you'll get more benefit from spending a little more money for decent kit.