<HTML dir=ltr><HEAD><TITLE>Re: [Mauiusers] nodeallocationpolicy</TITLE>
<META http-equiv=Content-Type content="text/html; charset=unicode">
<META content="MSHTML 6.00.2900.3395" name=GENERATOR></HEAD>
<BODY>
<DIV id=idOWAReplyText46464 dir=ltr>
<DIV dir=ltr><FONT face=Arial color=#000000 size=2>on my system, i use JOBNODEMATCHPOLICY EXACTNODE to ensure that users get specifically the resources they request (i.e. node=4:ppn=1 will give 4 nodes with 1 processor per node, no matter how many processors are on each system).</FONT></DIV>
<DIV dir=ltr><FONT face=Arial size=2></FONT> </DIV>
<DIV dir=ltr><FONT face=Arial size=2>--Joe</FONT></DIV></DIV>
<DIV dir=ltr><BR>
<HR tabIndex=-1>
<FONT face=Tahoma size=2><B>From:</B> Lawrence Sorrillo [mailto:sorrillo@jlab.org]<BR><B>Sent:</B> Wed 9/10/2008 10:52 AM<BR><B>To:</B> Greenseid, Joseph M.<BR><B>Cc:</B> A.Th.C. Hulst; mauiusers@supercluster.org<BR><B>Subject:</B> Re: [Mauiusers] nodeallocationpolicy<BR></FONT><BR></DIV>
<DIV>
<P><FONT size=2>Indeed how is this done?<BR><BR>The job should run on only 4 of the 8 cpus available per node and be<BR>spread across 4 nodes. So yes, using 16 cpus.<BR><BR>-L<BR><BR><BR>Greenseid, Joseph M. wrote:<BR>> wait, you want 16 CPU, divided evenly over 4 nodes, is that right?<BR>> <BR>> --Joe<BR>><BR>> *From:* mauiusers-bounces@supercluster.org on behalf of A.Th.C. Hulst<BR>> *Sent:* Mon 9/8/2008 7:41 AM<BR>> *To:* mauiusers@supercluster.org<BR>> *Subject:* Re: [Mauiusers] nodeallocationpolicy<BR>><BR>> Ah, sorry, I've found it. With moab one can request nodes=4:tpn=4,<BR>> which does<BR>> what I expect. I'll need moab if I want to try this.<BR>><BR>> Thanks<BR>><BR>> On Monday 08 September 2008 13:22:46 A.Th.C. Hulst wrote:<BR>> > #PBS -l nodes=4:ppn=4<BR>> > lamboot<BR>> > mpiexec hello_mpi<BR>> > lamhalt<BR>> ><BR>> > What happens is that the job is run on two nodes, each node running two<BR>> > chunks of 4 "cpus" (which are actually cores). That is not really<BR>> > desirable. I would like to divide the job over 4 nodes.<BR>> ><BR>> > My initial guess was that I could control node allocation with the<BR>> > nodeallocationpolicy parameter, but I fail to get it working.<BR>> ><BR>> > Does anyone have any experience with a similar issue? Is it possible?<BR>> ><BR>> > Best regards,<BR>> > Sander<BR>> ><BR>> > _______________________________________________<BR>> > mauiusers mailing list<BR>> > mauiusers@supercluster.org<BR>> > <A href="http://www.supercluster.org/mailman/listinfo/mauiusers">http://www.supercluster.org/mailman/listinfo/mauiusers</A><BR>><BR>><BR>><BR>> _______________________________________________<BR>> mauiusers mailing list<BR>> mauiusers@supercluster.org<BR>> <A href="http://www.supercluster.org/mailman/listinfo/mauiusers">http://www.supercluster.org/mailman/listinfo/mauiusers</A><BR>><BR>> ------------------------------------------------------------------------<BR>><BR>> _______________________________________________<BR>> mauiusers mailing list<BR>> mauiusers@supercluster.org<BR>> <A href="http://www.supercluster.org/mailman/listinfo/mauiusers">http://www.supercluster.org/mailman/listinfo/mauiusers</A><BR>> <BR><BR><BR></FONT></P></DIV></BODY></HTML>