Re: 2.6.5-mm5

[Nick Piggin]

markw@xxxxxxxx wrote:
> I have more results with DBT-2 on my 4-way Xeon system:
> 	http://developer.osdl.org/markw/fs/dbt2_project_results.html
>
> It doesn't look like the latest cpu scheduler work is helping this
> workload.  I've also made sure that the database was set to use fsync
> instead of fdatasync so you can see if those fsync speedup patches are
> offering anything with this workload too.
>
>            ext2  ext3
> 2.6.5-mm5  2165  1933
> 2.6.5-mm4  2180
> 2.6.5-mm3  2165  1930
> 2.6.5      2385
>
> Mark

Hmm, well the sched-less-idle patch is in mm5, which brought
2.6.5-rc3-mm4 to 2320 on ext2.

The only other significant scheduler changes since that kernel
are in -mm5.
+sched_less_idle
+sched_balance_context

So either sched_balance_context is causing a regression that
counters sched_less_idle, or maybe it isn't a scheduler problem?
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/