Chuck Lever wrote:
> if your stripe unit is small, then the disks will end up doing a bunch of
> small reads. if the stripe size is large, then the "small reads" are
> automatically coalesced into a large read. and if the read is aligned
> with the track cache, it is more likely that the disk's own read-ahead
> and caching systems will be able to work efficiently.
>
> that's the theory, anyway. your best bet is to get a file system
> benchmark you trust, and that mimics your application's disk access
> patterns, and use that to measure which configuration is best.
Some cursory experiments on this list seem to have yielded better large file
throughput numbers for stripe sizes of smaller amounts, say 4k. I tested a
large number of ide disks in various configurations of raid0 and raid5, and
my tests seemed to indicate the same thing... 4k chunk was faster than 128k
chunk for accessing large streaming files. None of these tests have been
particularly exhaustive or scientific, though.
I understand that this doesn't necessarily fit in with the theory of these
things. However, current raid performance may not match theory in a number
of places... we have limited real world advice to be given.
Would anyone here be up for participating in some more exhaustive benchmarks
for the good of the list? I can do some more scientific tests, and I'd be
happy to coordinate the effort. The result might allow us to produce a good
answer for this frequently asked question...as well as some others, and
might be able to provide good information to Ingo in terms of the state of
performance of raid.
Tom
