Test results from the OSDL reaim test.
The -mm kernels now appear to be scaling a bit nice.
I dunno why, but the 8-ways like -mm2 :)
The is the 'database' load, a mixture of IO and CPU activity.
2-CPU - (all AS scheduler)
Kernel Max JPM Percent change
linux-2.6.3 1266.95 0.0
2.6.4-rc1 1284.43 1.38
2.6.4-rc1-mm2 1316.93 3.90
4-CPU ( all AS )
linux-2.6.3 5313.36 0.0
2.6.4-rc1 5218.87 -1.78
2.6.4-rc1-mm2 5391.00 1.46
8-CPU (both)
linux-2.6.3 8663,87 0.0 (deadline)
linux-2.6.3 8776.32 1.35 (AS)
2.6.4-rc1 8664.95 0.0 (deadline)
2.6.4-rc1 8795.75 1.57 (AS)
2.6.4-rc1-mm2 9405.53 8.62 (deadline)
2.6.4-rc1-mm2 9159.24 5.77 (AS)
--------
cliffw
OSDL
http:// developer.osdl.org/cliffw/reaim ( More results )
--
The church is near, but the road is icy.
The bar is far, but i will walk carefully. - Russian proverb
cliff white <[email protected]> wrote:
>
>
> Test results from the OSDL reaim test.
> The -mm kernels now appear to be scaling a bit nice.
> I dunno why, but the 8-ways like -mm2 :)
I think your'e playing with my mind.
> The is the 'database' load, a mixture of IO and CPU activity.
What about file server load?
> 4-CPU ( all AS )
> linux-2.6.3 5313.36 0.0
> 2.6.4-rc1 5218.87 -1.78
> 2.6.4-rc1-mm2 5391.00 1.46
I spent a boring evening with the file server load on 4-way x86 with six
disks. If I squinted at it hard enough I was able to discern a 1% slowdown
due to O_DIRECT-vs-buffered-fix.patch, but it was pretty thin.
I didn't test the database load.
On Mon, 8 Mar 2004 14:40:50 -0800
Andrew Morton <[email protected]> wrote:
> cliff white <[email protected]> wrote:
> >
> >
> > Test results from the OSDL reaim test.
> > The -mm kernels now appear to be scaling a bit nice.
> > I dunno why, but the 8-ways like -mm2 :)
>
> I think your'e playing with my mind.
It's not _me, it's those STP robots. :)
>
> > The is the 'database' load, a mixture of IO and CPU activity.
>
> What about file server load?
Also looking pretty okay on ext3, and sucking somewhat less
on the other filesystems.
I run filesystem compares on two-ways, so the
delta is not as big.
2-CPU-
linux-2.6.3 6055.26 0.0 ext3
2.6.4-rc1-mm2 6149.28 1.64 ext3
2.6.4-rc1 6004.18 -0.84 ext3
2.6.4-rc1-mm2 6061.28 0.1 jfs
2.6.4-rc2 5977.53 -1.36 jfs
2.6.4-rc1-mm2 6007.67 -0.83 reiserfs
2.6.4-rc1 5836.39 -3.83 reiserfs
2.6.4-rc1-mm2 5801.30 -4.44 xfs
2.6.4-rc1 5785.93 -4.71 xfs
----
>
> > 4-CPU ( all AS )
> > linux-2.6.3 5313.36 0.0
> > 2.6.4-rc1 5218.87 -1.78
> > 2.6.4-rc1-mm2 5391.00 1.46
>
> I spent a boring evening with the file server load on 4-way x86 with six
> disks. If I squinted at it hard enough I was able to discern a 1% slowdown
> due to O_DIRECT-vs-buffered-fix.patch, but it was pretty thin.
Gee, i told the STP robots to run the tests, and went out and had fun :)
We should talk, the robots would be your slaves, if you would but ask them.
>
> I didn't test the database load.
The db load mixes the IO with some compute stuff, it will always run faster
than the fserver load.
cliffw
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
--
The church is near, but the road is icy.
The bar is far, but i will walk carefully. - Russian proverb