> backwards this time. dd if=/dev/zero of=/dev/rd/c0d{0,1} bs=512 count=100
>
> what's /dev/rd/? devfs or a RAID controller?
it's the raid controller (Mylex DAC1164P). First time I've dealt with
something that put its devices there :)
> are you running the bonnies in singleuser? if so cpu util will nearly
> always approach 100%.
runlevel 3, normal multiuser, stock RH 6.0 install (besides kernel
which is 2.2.11-ac3). I pass mem=1024M since bonnie on ext2 has a
signed int (2^31-1) byte limitation and the 2GB of ram in the machine
I had the "memory support" option at 2GB since I was going to want to
use it all after bonnie runs, but I'll turn it back to 1GB as I turn
the kernel back to UP for testing.
> could you try building a uni-processor kernel and running your tests
> on that? i'm wondering about smp overhead...
UP kernel results (s/w 0 over 2 h/w 0's)
-------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
2047 22378 97.8 45738 32.6 20770 42.1 26389 96.0 44574 58.4 569.2 6.4
compared the 4-way Xeon numbers of:
-------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
2047 22238 99.0 54198 43.2 20813 47.0 24282 90.4 42598 60.5 623.5 7.3
For writing, CPU util went DOWN, rates got a little worse (block).
For reading, CPU util stayed about the same, rates got BETTER.
Either bonnie prints out wildly wrong numbers for CPU util, or adding
in the use of the 3 Xeon's was cancelled out by SMP overhead (?)
Just to make sure the numbers weren't caused by dropping from 2GB to
1GB in the memory support section, I re-ran 4-way with 1GB support:
-------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
2047 22103 99.6 55974 44.9 20778 46.2 23958 89.2 42752 61.2 591.2 4.6
These numbers nearly match the previous 4-way numbers (slight variance
at best) so the 1GB/2GB selection doesn't appear to matter.
James
--
Miscellaneous Engineer --- IBM Netfinity Performance Development
