[torqueusers] cpusets in Torque 3.0, memory pressure

David Beer dbeer at adaptivecomputing.com
Thu Jun 7 09:51:14 MDT 2012


On Wed, Jun 6, 2012 at 10:44 PM, Grigory Shamov <gas5x at yahoo.com> wrote:

> Hi David,
>
> Thanks!
>
>  I've tried to run some chemistry codes,  under Torque 3, with and without
> oversubscription for memory. It seems to work. The pressure  can quickly
> spike up to, say, 13000 with quite moderate swapping.
>
> If there are more than one jobs on the node, and one of them
> oversubscribes, would it create pressure for all the jobs there?
>

I don't think that it would, but I would verify that experimentally.

David


>
> --
> Grigory Shamov
>
>
>
> --- On *Mon, 6/4/12, David Beer <dbeer at adaptivecomputing.com>* wrote:
>
>
> From: David Beer <dbeer at adaptivecomputing.com>
> Subject: Re: [torqueusers] cpusets in Torque 3.0, memory pressure
> To: "Torque Users Mailing List" <torqueusers at supercluster.org>
> Date: Monday, June 4, 2012, 7:27 AM
>
>
>
>
> On Sun, Jun 3, 2012 at 7:47 AM, Grigory Shamov <gas5x at yahoo.com<http://mc/compose?to=gas5x@yahoo.com>
> > wrote:
>
> Dear David,
>
>
> Thanks a lot for your  answer!
>
>
> What would be a good starting value for the memory pressure threshold, to
> configure it right?
>
>
> I think 1000 is a good starting value.
>
> David
>
>
>
> --
>
> Grigory Shamov
>
>
>
>
> --- On *Fri, 6/1/12, David Beer <dbeer at adaptivecomputing.com<http://mc/compose?to=dbeer@adaptivecomputing.com>
> >* wrote:
>
>
> From: David Beer <dbeer at adaptivecomputing.com<http://mc/compose?to=dbeer@adaptivecomputing.com>
> >
> Subject: Re: [torqueusers] cpusets in Torque 3.0, memory pressure
> To: "Torque Users Mailing List" <torqueusers at supercluster.org<http://mc/compose?to=torqueusers@supercluster.org>
> >
> Date: Friday, June 1, 2012, 1:07 PM
>
>
>
>
> On Fri, Jun 1, 2012 at 11:57 AM, Grigory Shamov <gas5x at yahoo.com<http://mc/compose?to=gas5x@yahoo.com>
> > wrote:
>
> Hi,
>
> The Torque 3 changelog says " when cpusets are configured and memory
> pressure enabled, add the ability to check memory pressure for a job. Using
> $memory_pressure_threshold and $memory_pressure_duration in the mom's
> config, the admin sets a threshold at  which a job becomes a problem".
>
> Does it work for non-NUMA setup as well? I mean, if I configure it with
> --enable-cpusets only, but --disable-numa-support ?
>
>
> Grigory,
>
> As long as you set your cpuset up correctly (
> http://www.adaptivecomputing.com/resources/docs/torque/3-0-3/a.cmomconfig.php)
> it will work with or without numa support.
>
> David
>
>
> --
> Grigory Shamov
> HPC Analyst,
> University of Manitoba
>
>
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org<http://mc/compose?to=torqueusers@supercluster.org>
> http://www.supercluster.org/mailman/listinfo/torqueusers
>
>
>
>
> --
> David Beer | Software Engineer
> Adaptive Computing
>
>
> -----Inline Attachment Follows-----
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org<http://mc/compose?to=torqueusers@supercluster.org>
> http://www.supercluster.org/mailman/listinfo/torqueusers
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org<http://mc/compose?to=torqueusers@supercluster.org>
> http://www.supercluster.org/mailman/listinfo/torqueusers
>
>
>
>
> --
> David Beer | Software Engineer
> Adaptive Computing
>
>
> -----Inline Attachment Follows-----
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org<http://mc/compose?to=torqueusers@supercluster.org>
> http://www.supercluster.org/mailman/listinfo/torqueusers
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>
>


-- 
David Beer | Software Engineer
Adaptive Computing
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torqueusers/attachments/20120607/218b7388/attachment-0001.html 


More information about the torqueusers mailing list