Re: Lockstats for SMP DB Workload

Hanna Linder (hannal@us.ibm.com)
Tue, 04 Jun 2002 18:10:25 -0700


--On Tuesday, June 04, 2002 17:16:46 -0700 Ruth Forester <lilo@us.ibm.com> wrote:

> Everyone,
>
> I am running with the following configuration
>
> 2.4.19pre8aa2+dj2+ (dj2 removes global semaphore_lock spinlock)
> +fast_walkA3-2_4_19-pre8_patch.
>
> The database is set up to use raw-io, yet looking at this data, it appears that
> I am still hitting a lot of filesystem accesses, among other things. This is an
> oltp workload, although there are some contentions (pread?) that cause the
> cpu sys time to go to 99%, it was during this part of the "workload" that this
> snapshot of lockmeter was taken.
>

> SPINLOCKS HOLD WAIT
> UTIL CON MEAN( MAX ) MEAN( MAX )(% CPU) TOTAL NOWAIT SPIN RJECT NAME
>
> 6.7% 2.5us( 25ms) 62us( 12ms)( 6.8%) 2650959 93.3% 6.6% 0.05% *TOTAL*
>
> 2.4% 1.0% 38us( 15ms) 3185us( 12ms)(0.13%) 6291 99.0% 1.0% 0% kernel_flag_cacheline
> 2.3% 9.1% 6865us( 15ms) 3551us(5592us)(0.01%) 33 90.9% 9.1% 0% do_exit+0xf4

Dave Hanson has a do_exit patch you might want to look at...

Hanna

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/