<br><br><div class="gmail_quote">On Wed, Jun 6, 2012 at 10:44 PM, Grigory Shamov <span dir="ltr">&lt;<a href="mailto:gas5x@yahoo.com" target="_blank">gas5x@yahoo.com</a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<table cellspacing="0" cellpadding="0" border="0"><tbody><tr><td valign="top" style="font:inherit">Hi David,<br><br>Thanks!<br><br> I&#39;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.<br>
<br>If there are more than one jobs on the node, and one of them oversubscribes, would it create pressure for all the jobs there?<br></td></tr></tbody></table></blockquote><div><br></div><div>I don&#39;t think that it would, but I would verify that experimentally.</div>
<div><br></div><div>David</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><table cellspacing="0" cellpadding="0" border="0"><tbody><tr><td valign="top" style="font:inherit">
<br>--<br>Grigory Shamov<div class="im"><br><br><br>--- On <b>Mon, 6/4/12, David Beer <i>&lt;<a href="mailto:dbeer@adaptivecomputing.com" target="_blank">dbeer@adaptivecomputing.com</a>&gt;</i></b> wrote:<br></div><blockquote style="border-left:2px solid rgb(16,16,255);margin-left:5px;padding-left:5px">
<div class="im"><br>From: David Beer &lt;<a href="mailto:dbeer@adaptivecomputing.com" target="_blank">dbeer@adaptivecomputing.com</a>&gt;<br>Subject: Re: [torqueusers] cpusets in Torque 3.0, memory pressure<br>To: &quot;Torque Users Mailing List&quot; &lt;<a href="mailto:torqueusers@supercluster.org" target="_blank">torqueusers@supercluster.org</a>&gt;<br>
</div>Date: Monday, June 4, 2012, 7:27 AM<div><div class="h5"><br><br><div><br><br><div>On Sun, Jun 3, 2012 at 7:47 AM, Grigory Shamov <span dir="ltr">&lt;<a rel="nofollow" href="http://mc/compose?to=gas5x@yahoo.com" target="_blank">gas5x@yahoo.com</a>&gt;</span> wrote:<br>
<blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<table border="0" cellpadding="0" cellspacing="0"><tbody><tr><td style="font:inherit" valign="top"><p>Dear David,<br></p><p><br></p><p>Thanks a lot for your  answer! </p><p><br></p><p>What would be a good starting value for the memory pressure threshold, to configure it right? </p>

</td></tr></tbody></table></blockquote><div><br>I think 1000 is a good starting value.<br><br>David<br> </div><blockquote style="margin:0pt 0pt 0pt 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<table border="0" cellpadding="0" cellspacing="0"><tbody><tr><td style="font:inherit" valign="top"><p><br></p><p>--</p><p>Grigory Shamov</p><p><br></p><p><br></p><br>--- On <b>Fri, 6/1/12, David Beer <i>&lt;<a rel="nofollow" href="http://mc/compose?to=dbeer@adaptivecomputing.com" target="_blank">dbeer@adaptivecomputing.com</a>&gt;</i></b> wrote:<br>

<blockquote style="border-left:2px solid rgb(16,16,255);margin-left:5px;padding-left:5px"><br>From: David Beer &lt;<a rel="nofollow" href="http://mc/compose?to=dbeer@adaptivecomputing.com" target="_blank">dbeer@adaptivecomputing.com</a>&gt;<br>
Subject: Re: [torqueusers] cpusets in Torque 3.0, memory pressure<br>
To: &quot;Torque Users Mailing List&quot; &lt;<a rel="nofollow" href="http://mc/compose?to=torqueusers@supercluster.org" target="_blank">torqueusers@supercluster.org</a>&gt;<br>Date: Friday, June 1, 2012, 1:07 PM<div><div>
<br><br><div><br>
<br><div>On Fri, Jun 1, 2012 at 11:57 AM, Grigory Shamov <span dir="ltr">&lt;<a rel="nofollow" href="http://mc/compose?to=gas5x@yahoo.com" target="_blank">gas5x@yahoo.com</a>&gt;</span> wrote:<br><blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">


Hi,<br>
<br>
The Torque 3 changelog says &quot; 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&#39;s config, the admin sets a threshold at  which a job becomes a problem&quot;.<br>



<br>
Does it work for non-NUMA setup as well? I mean, if I configure it with --enable-cpusets only, but --disable-numa-support ?<br>
<br></blockquote><div><br></div><div>Grigory,</div><div><br></div><div>As long as you set your cpuset up correctly (<a rel="nofollow" href="http://www.adaptivecomputing.com/resources/docs/torque/3-0-3/a.cmomconfig.php" target="_blank">http://www.adaptivecomputing.com/resources/docs/torque/3-0-3/a.cmomconfig.php</a>) it will work with or without numa support.</div>


<div><br></div><div>David</div><div> </div><blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
--<br>
Grigory Shamov<br>
HPC Analyst,<br>
University of Manitoba<br>
<br>
<br>
<br>
<br>
_______________________________________________<br>
torqueusers mailing list<br>
<a rel="nofollow" href="http://mc/compose?to=torqueusers@supercluster.org" target="_blank">torqueusers@supercluster.org</a><br>
<a rel="nofollow" href="http://www.supercluster.org/mailman/listinfo/torqueusers" target="_blank">http://www.supercluster.org/mailman/listinfo/torqueusers</a><br>
</blockquote></div><br><br clear="all"><div><br></div>-- <br><div>David Beer | Software Engineer</div><div>Adaptive Computing</div><br>
</div><br></div></div>-----Inline Attachment Follows-----<div><br><br><div>_______________________________________________<br>torqueusers mailing list<br><a rel="nofollow" href="http://mc/compose?to=torqueusers@supercluster.org" target="_blank">torqueusers@supercluster.org</a><br>

<a rel="nofollow" href="http://www.supercluster.org/mailman/listinfo/torqueusers" target="_blank">http://www.supercluster.org/mailman/listinfo/torqueusers</a><br></div></div></blockquote></td></tr></tbody></table><br>_______________________________________________<br>


torqueusers mailing list<br>
<a rel="nofollow" href="http://mc/compose?to=torqueusers@supercluster.org" target="_blank">torqueusers@supercluster.org</a><br>
<a rel="nofollow" href="http://www.supercluster.org/mailman/listinfo/torqueusers" target="_blank">http://www.supercluster.org/mailman/listinfo/torqueusers</a><br>
<br></blockquote></div><br><br clear="all"><br>-- <br><div>David Beer | Software Engineer</div><div>Adaptive Computing</div><br>
</div><br>-----Inline Attachment Follows-----<br><br><div>_______________________________________________<br>torqueusers mailing list<br><a href="http://mc/compose?to=torqueusers@supercluster.org" target="_blank">torqueusers@supercluster.org</a><br>
<a href="http://www.supercluster.org/mailman/listinfo/torqueusers" target="_blank">http://www.supercluster.org/mailman/listinfo/torqueusers</a><br></div></div></div></blockquote></td></tr></tbody></table><br>_______________________________________________<br>

torqueusers mailing list<br>
<a href="mailto:torqueusers@supercluster.org">torqueusers@supercluster.org</a><br>
<a href="http://www.supercluster.org/mailman/listinfo/torqueusers" target="_blank">http://www.supercluster.org/mailman/listinfo/torqueusers</a><br>
<br></blockquote></div><br><br clear="all"><div><br></div>-- <br><div>David Beer | Software Engineer</div><div>Adaptive Computing</div><br>