[Beowulf] C-States and Latency

[Kevin Van Workum]

On Mon, Feb 24, 2014 at 10:58 PM, Christopher Samuel
<samuel at unimelb.edu.au> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 21/02/14 09:23, Kevin Van Workum wrote:
>
>> Does anyone have a real-world application where setting
>> intel_idle.max_cstate=0 and processor.max_cstate=0 actually makes a
>> difference wrt application or network latency?
>
> Mellanox's recommendation is:
>
> intel_idle.max_cstate=0 processor.max_cstate=1
>
> That's from:
>
> http://www.mellanox.com/related-docs/prod_software/Performance_Tuning_Guide_for_Mellanox_Network_Adapters.pdf
>
> We certainly saw performance problems with RHEL6 on Sandybridge systems
> without that.
>

Yes, this is the recommendation I'm wondering about. Can you describe
in more detail the performance problems you saw? Any particular
application? Measurables?

--