Re: [PATCH 0/3][RFC] Improve load balancing when tasks have largeweight differential

[Nikhil Rao]

On Tue, Sep 28, 2010 at 6:57 AM, Mike Galbraith <efault@xxxxxx> wrote:
> On Mon, 2010-09-27 at 17:29 -0700, Nikhil Rao wrote:
>> Hi all,
>>
>> I have attached a series of patches that improve load balancing when there is a
>> large weight differential between tasks. These patches are based off the
>> feedback Peter Zijlstra gave in an earlier post (see http://thread.gmane.org/gmane.linux.kernel/1015966).
>> They can be applied to v2.6.36-rc5 or -tip without conflicts.
>>
>> Tested with the following setup.
>> - Test machine is a 16 cpu box (quad-socket, quad-core).
>> - Baseline is v2.6.36-rc5 kernel
>>
>> We spawn 16 SCHED_IDLE soaker threads and one SCHED_NORMAL task. On the
>> baseline kernel, the machine has ~18% idle time. With these patches applied on
>> top of baseline, idle time drops to 0%.
>
> Hm. I can get it stuck with one core idle on ym little quad.
>
> top - 15:53:22 up 11 min, 17 users, Âload average: 5.05, 4.40, 2.51
> Tasks: 270 total, Â 7 running, 263 sleeping, Â 0 stopped, Â 0 zombie
> Cpu(s): 75.3%us, Â0.0%sy, Â0.0%ni, 24.7%id, Â0.0%wa, Â0.0%hi, Â0.0%si, Â0.0%st
>
> ÂPID USER Â Â ÂPR ÂNI ÂVIRT ÂRES ÂSHR S %CPU %MEM Â ÂTIME+ ÂP COMMAND
> Â7455 root    5 -15 Â7996 Â340 Â256 R Â100 Â0.0  0:59.93 1 pert
> Â7421 root   Â20  0 Â7996 Â340 Â256 R  50 Â0.0  4:20.01 3 pert
> Â7422 root   Â20  0 Â7996 Â340 Â256 R  50 Â0.0  3:45.81 2 pert
> Â7423 root   Â20  0 Â7996 Â340 Â256 R  50 Â0.0  4:09.45 2 pert
> Â7424 root   Â20  0 Â7996 Â344 Â256 R  50 Â0.0  4:12.75 3 pert
>
>

Mike,

Thanks for running this. I've not been able to reproduce what you are
seeing on the few test machines that I have (different combinations of
MC, CPU and NODE domains). Can you please give me more info about
your setup?

-Thanks,
Nikhil
--
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/