<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic Hi Dan, in Intel® MPI Library</title>
    <link>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029207#M4151</link>
    <description>&lt;P&gt;Hi Dan,&lt;/P&gt;

&lt;P&gt;ITAC&amp;nbsp;is not aware about File System. It uses regular files (FILE *) to store traces.&lt;BR /&gt;
	But, it seems to me that the issue is related to the algorithm of saving files&amp;nbsp; - only rank0 saves traces and it can be a bottleneck.&lt;/P&gt;

&lt;P&gt;Regards!&lt;BR /&gt;
	---Dmitry&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;</description>
    <pubDate>Fri, 23 Oct 2015 07:52:51 GMT</pubDate>
    <dc:creator>Dmitry_K_Intel2</dc:creator>
    <dc:date>2015-10-23T07:52:51Z</dc:date>
    <item>
      <title>ITAC Lustre integration</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029206#M4150</link>
      <description>&lt;P&gt;I am using itac-9.0.3.051 on a Linux cluster to profile rank 0 and rank 15 of a small MPI application.&amp;nbsp; The code was compiled with the Intel compiler suite (2015.3.187) and impi-5.0.3.048.&amp;nbsp; The compile flags include "-g -tcollect" and I started the MPI with "mpirun -trace".&amp;nbsp; The application took a VERY long time to complete, but it did finish.&amp;nbsp; Now ITAC is writing the trace data (~2328778.00 MB) to our Lustre file system.&amp;nbsp; The problem is that I'm only getting about 50MB/s to this file system which is capable of much higher write speeds..&amp;nbsp; Does ITAC have any internal awareness of Lustre like Intel MPI does?&amp;nbsp; What ITAC settings would you suggest I use to achieve optimal writes to Lustre?&lt;/P&gt;

&lt;P&gt;Here are my ITAC settings&lt;/P&gt;

&lt;P&gt;setenv VT_CONFIG ${PBS_O_WORKDIR}/itac.conf&lt;BR /&gt;
	setenv VT_LOGFILE_PREFIX ${PBS_O_WORKDIR}/ITAC.${jobid}&lt;BR /&gt;
	setenv VT_FLUSH_PREFIX $VT_LOGFILE_PREFIX&lt;BR /&gt;
	setenv VT_MEM_BLOCKSIZE 2097152&lt;BR /&gt;
	setenv VT_MEM_FLUSHBLOCKS 32&lt;BR /&gt;
	setenv VT_MEM_MAXBLOCKS 512&lt;BR /&gt;
	mkdir $VT_LOGFILE_PREFIX&lt;BR /&gt;
	lfs setstripe -c 4 -s 2097152 -i -1 $VT_LOGFILE_PREFIX&lt;/P&gt;

&lt;P&gt;Dan&lt;/P&gt;</description>
      <pubDate>Thu, 22 Oct 2015 21:24:00 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029206#M4150</guid>
      <dc:creator>dkokron</dc:creator>
      <dc:date>2015-10-22T21:24:00Z</dc:date>
    </item>
    <item>
      <title>Hi Dan,</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029207#M4151</link>
      <description>&lt;P&gt;Hi Dan,&lt;/P&gt;

&lt;P&gt;ITAC&amp;nbsp;is not aware about File System. It uses regular files (FILE *) to store traces.&lt;BR /&gt;
	But, it seems to me that the issue is related to the algorithm of saving files&amp;nbsp; - only rank0 saves traces and it can be a bottleneck.&lt;/P&gt;

&lt;P&gt;Regards!&lt;BR /&gt;
	---Dmitry&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;</description>
      <pubDate>Fri, 23 Oct 2015 07:52:51 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029207#M4151</guid>
      <dc:creator>Dmitry_K_Intel2</dc:creator>
      <dc:date>2015-10-23T07:52:51Z</dc:date>
    </item>
    <item>
      <title>As both Intel MPI and Lustre</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029208#M4152</link>
      <description>&lt;P&gt;As both Intel MPI and Lustre are Intel products, please consider collaborating with those teams to add parallel I/O and Lustre awareness to ITAC.&lt;/P&gt;</description>
      <pubDate>Fri, 23 Oct 2015 13:40:34 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/ITAC-Lustre-integration/m-p/1029208#M4152</guid>
      <dc:creator>dkokron</dc:creator>
      <dc:date>2015-10-23T13:40:34Z</dc:date>
    </item>
  </channel>
</rss>

