[ Wednesday, March 1, 2000 ] Ricky Beam wrote:
> > -------Sequential Output-------- ---Sequential Input-- --Random--
> > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
> > MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
> > 256 5451 78.7 10035 8.5 4000 7.3 3975 55.3 18765 11.3 262.8 3.9
>
> That's a lot of CPU being used for a hardware RAID device.
per-char doesn't matter (one of the reasons I hate ppl using bonnie,
besides the single-threaded-ness). Considering the queueing and scat/gat
the driver is probably trying (not to mention caching, esp. in the read
case), 7.3-11.3% seems acceptable for the block stuff.
I need to check to see if madvise() has been backported to 2.2.x, as
MADV_RANDOM may help cut down or eliminate memory caching effects... it'd
be nice to get (approx) the same numbers from 100MB and 1000MB test runs,
regardless of the amount of memory in the machines :)
James
