<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    Hi Abu,<br>
    <br>
    I think Jeff means that you should run your experiment with more
    iterations in order to get a stable results.<br>
    - Increase the iteration of for loop in each execution (I think osu
    benchmark allows you to set it)<br>
    - Run the experiments 10 or 100 times, and take the average and
    standard deviation.<br>
    <br>
    If you see a very small standard deviation (e.g., <=5%), then the
    trend is stable and you might not see such gaps.<br>
    <br>
    Best regards,<br>
    Min<br>
    <div class="moz-cite-prefix">On 2018/06/20 12:14, Abu Naser wrote:<br>
    </div>
    <blockquote type="cite" cite="mid:BLUPR0501MB200383DFBCA0721E71DDAC0E97770@BLUPR0501MB2003.namprd05.prod.outlook.com">
      
      <style type="text/css" style="display:none;"><!-- P {margin-top:0;margin-bottom:0;} --></style>
      <div id="divtagdefaultwrapper" style="font-size:12pt;color:#000000;font-family:Calibri,Helvetica,sans-serif;" dir="ltr">
        <p style="margin-top:0;margin-bottom:0">Hello Jeff,</p>
        <p style="margin-top:0;margin-bottom:0"><br>
        </p>
        <p style="margin-top:0;margin-bottom:0">Yes, I am using a switch
          and other machines are also connected with that switch.
          <br>
        </p>
        <p style="margin-top:0;margin-bottom:0">If I remove other
          machines and just use my two node with the switch, then will
          it improve the performance by 200 ~ 400 iterations?</p>
        <p style="margin-top:0;margin-bottom:0">Meanwhile I will give a
          try with a single dedicated cable.
          <span></span><br>
        </p>
        <p style="margin-top:0;margin-bottom:0"><br>
        </p>
        <p style="margin-top:0;margin-bottom:0">Thank you.<br>
        </p>
        <div id="Signature">
          <div id="divtagdefaultwrapper" dir="ltr" style="font-size:12pt; color:rgb(0,0,0);
font-family:Calibri,Helvetica,sans-serif,"EmojiFont","Apple
            Color Emoji","Segoe UI
            Emoji",NotoColorEmoji,"Segoe UI
            Symbol","Android Emoji",EmojiSymbols">
            <p><br>
            </p>
            <p align="left"><span style="font-size:10pt;
                font-family:Calibri,Helvetica,sans-serif">Best Regards,</span></p>
            <span style="font-family:Calibri,Helvetica,sans-serif;
              font-size:10pt"></span>
            <div align="left"><span style="font-size:11pt;
                font-family:Calibri,Helvetica,sans-serif"></span></div>
            <span style="font-family:Calibri,Helvetica,sans-serif;
              font-size:10pt"></span>
            <p align="left"><span style="font-size:10pt;
                font-family:Calibri,Helvetica,sans-serif">Abu Naser</span><br>
            </p>
          </div>
        </div>
      </div>
      <hr style="display:inline-block;width:98%" tabindex="-1">
      <div id="divRplyFwdMsg" dir="ltr"><font style="font-size:11pt" face="Calibri, sans-serif" color="#000000"><b>From:</b> Jeff
          Hammond <a class="moz-txt-link-rfc2396E" href="mailto:jeff.science@gmail.com"><jeff.science@gmail.com></a><br>
          <b>Sent:</b> Wednesday, June 20, 2018 12:52:06 PM<br>
          <b>To:</b> MPICH<br>
          <b>Subject:</b> Re: [mpich-discuss] osu_latency test: why 8KB
          takes less time than 4KB and 2KB takes less time than 1KB?</font>
        <div> </div>
      </div>
      <meta content="text/html; charset=utf-8">
      <div>
        <div dir="ltr">Is the ethernet connection a single dedicated
          cable between the two machines or are you running through a
          switch that handles other traffic?
          <div><br>
          </div>
          <div>My best guess is that this is noise and that you may be
            able to avoid it by running a very long time, e.g. 10000
            iterations.</div>
          <div><br>
          </div>
          <div>Jeff</div>
        </div>
        <div class="x_gmail_extra"><br>
          <div class="x_gmail_quote">On Wed, Jun 20, 2018 at 6:53 AM,
            Abu Naser <span dir="ltr">
              <<a href="mailto:an16e@my.fsu.edu" target="_blank" moz-do-not-send="true">an16e@my.fsu.edu</a>></span>
            wrote:<br>
            <blockquote class="x_gmail_quote" style="margin:0 0 0 .8ex;
              border-left:1px #ccc solid; padding-left:1ex">
              <div dir="ltr">
                <div id="x_m_6077755676379859201divtagdefaultwrapper" dir="ltr" style="font-size:12pt; color:rgb(0,0,0);
font-family:Calibri,Helvetica,sans-serif,"EmojiFont","Apple
                  Color Emoji","Segoe UI
                  Emoji",NotoColorEmoji,"Segoe UI
                  Symbol","Android Emoji",EmojiSymbols">
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">Good day to
                    all,</p>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">I had run
                    point to point osu_latency test in two nodes for 200
                    times.  Followings are the average time in
                    microsecond for various size of the messages -</p>
                  <div>1KB    84.8514 us<br>
                    <span style="color:rgb(255,0,0)">2KB    73.52535</span>
                    us<br>
                    4KB    272.55275 us<br>
                    <span style="color:rgb(255,0,0)">8KB    234.86385</span>
                    us<br>
                    16KB    288.88 us<br>
                    32KB    523.3725 us<br>
                    64KB    910.4025 us</div>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">From the
                    above looks like, 2KB message has less latency than
                    1 KB and 8KB has less latency than 4KB.
                    <br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">I was looking
                    for explanation of this behavior  but did not get
                    any.</p>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <ol style="margin-bottom:0px; margin-top:0px">
                    <li><span>MPIR_CVAR_CH3_EAGER_MAX_MSG_<wbr>SIZE</span><span>
                        is set to 128KB. So none of the above message
                        size is using Rendezvous protocol. Is there any
                        partition inside eager protocol (e.g. 0 - 512
                        bytes, 1KB - 8KB, 16KB - 64KB)? If yes then what
                        are the boundaries for them? Can I log them with
                        debug-event-logging? </span><br>
                    </li>
                  </ol>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">Setup I am
                    using:</p>
                  <p style="margin-top:0; margin-bottom:0">- two nodes
                    has intel core i7, one with 16gb memory another one
                    8gb</p>
                  <p style="margin-top:0; margin-bottom:0">- mpich
                    3.2.1, configured and build to use nemesis tcp</p>
                  <p style="margin-top:0; margin-bottom:0">- 1gb
                    Ethernet connection</p>
                  <p style="margin-top:0; margin-bottom:0">- NFS is
                    using for sharing<br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">- osu_latency
                    : uses MPI_Send and MPI_Recv</p>
                  <p style="margin-top:0; margin-bottom:0">- <span>MPIR_CVAR_CH3_EAGER_MAX_MSG_<wbr>SIZE</span>=
                    <span>131072</span> (128KB)<br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0">Can anyone
                    help me on that? Thanks in advance.<br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <p style="margin-top:0; margin-bottom:0"><br>
                  </p>
                  <div id="x_m_6077755676379859201Signature">
                    <div id="x_m_6077755676379859201divtagdefaultwrapper" dir="ltr" style="font-size:12pt; color:rgb(0,0,0);
font-family:Calibri,Helvetica,sans-serif,"EmojiFont","Apple
                      Color Emoji","Segoe UI
                      Emoji",NotoColorEmoji,"Segoe UI
                      Symbol","Android
                      Emoji",EmojiSymbols">
                      <p><br>
                      </p>
                      <p align="left"><span style="font-size:10pt;
                          font-family:Calibri,Helvetica,sans-serif">Best
                          Regards,</span></p>
                      <span style="font-family:Calibri,Helvetica,sans-serif;
                        font-size:10pt"></span>
                      <div align="left"><span style="font-size:11pt;
                          font-family:Calibri,Helvetica,sans-serif"></span></div>
                      <span style="font-family:Calibri,Helvetica,sans-serif;
                        font-size:10pt"></span>
                      <p align="left"><span style="font-size:10pt;
                          font-family:Calibri,Helvetica,sans-serif">Abu
                          Naser</span><br>
                      </p>
                    </div>
                  </div>
                </div>
              </div>
              <br>
              ______________________________<wbr>_________________<br>
              discuss mailing list     <a href="mailto:discuss@mpich.org" moz-do-not-send="true">discuss@mpich.org</a><br>
              To manage subscription options or unsubscribe:<br>
              <a href="https://lists.mpich.org/mailman/listinfo/discuss" rel="noreferrer" target="_blank" moz-do-not-send="true">https://lists.mpich.org/<wbr>mailman/listinfo/discuss</a><br>
              <br>
            </blockquote>
          </div>
          <br>
          <br clear="all">
          <div><br>
          </div>
          -- <br>
          <div class="x_gmail_signature">Jeff Hammond<br>
            <a href="mailto:jeff.science@gmail.com" target="_blank" moz-do-not-send="true">jeff.science@gmail.com</a><br>
            <a href="http://jeffhammond.github.io/" target="_blank" moz-do-not-send="true">http://jeffhammond.github.io/</a></div>
        </div>
      </div>
      <!--'"--><br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
discuss mailing list     <a class="moz-txt-link-abbreviated" href="mailto:discuss@mpich.org">discuss@mpich.org</a>
To manage subscription options or unsubscribe:
<a class="moz-txt-link-freetext" href="https://lists.mpich.org/mailman/listinfo/discuss">https://lists.mpich.org/mailman/listinfo/discuss</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>