> it also depends on the load. if bonnie is the only real load on the
> machine then cpu util will be 99%.
>
> james: does it affect responsiveness? and if you run a couple of
> other cpu hogging processes (eg rc5des at normal nice level), does
> bonnie's cpu util go down? and does it affect results.
I'm starting to doubt that bonnie's cpu util means anything and that
instead I'll need real kernel profiling for the s/w raid stuff...
I can't find anything in Configure.help or .config about it though,
and checking around for CONFIG_PROFILE only shows it up in the sparc
and mips branches... disturbing :)
Hmmmm ok, here's my best guess for hogs that should stay in their caches:
#include <pthread.h>
void *spin(void *junk) { while(1); }
int main() {
int i; pthread_t junk;
for(i=0;i<4;i++) pthread_create(&junk,NULL,spin,NULL);
sleep(600);
}
Side note: besides system("grep -c processor /proc/cpuinfo");
(or ^$ instead of processor), is there a better/simpler/more portable
across diff linux arch's method for fetching # of processors?
Fighting_with_hogs
-------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 17004 74.7 55532 43.8 19114 42.2 16770 59.8 39221 54.1 560.5 5.7
idle_machine
-------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
> my feeling is that cpu util will go down significantly, but bonnie
> performance will not.
Results: Block writes stayed about the same for perf and cpu util, while
block reads, both per-char, rewrite, and random seeks all got worse
perf (but also less cpu util).
So, can some kind soul help be figure out the cpu overhead involved
in the s/w raid sections?
James
--
Miscellaneous Engineer --- IBM Netfinity Performance Development
