<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p>One short section of farm performance got left out of the
      minutes. It is on the wiki now, also find it below.</p>
    <h3><span class="mw-headline" id="SciComp_System_Perfomance">SciComp
        System Perfomance</span></h3>
    <p>We got onto a digression on recent experience using the farm and
      disk resources at JLab.
    </p>
    <ul>
      <li> Richard reports short file life-times on the cache disk after
        retrieval from tape, not associated with farm jobs, as short as
        20 hours. We speculated that this was due to farm job usage of
        the cache displacing files that would normally live for a week.</li>
      <li> Random trigger files that are supposed to be "permanently"
        pinned for year have disappeared. This causes problems for Monte
        Carlo jobs on the OSG which need these files.</li>
      <li> We commented that although we have a lot of space on the work
        disk, there are usage storms on work that not only make file
        retrieval extremely slow, but bring the ifarm nodes to a
        standstill. When work was on Lustre this was not an issue, but
        handling small files was a problem.</li>
    </ul>
    <div class="moz-cite-prefix">On 3/29/19 2:08 PM, Mark Ito wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:f51d664d-62af-5fcb-1795-53849c678890@jlab.org">
      <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
      <p>Folks,</p>
      <p>Please find the minutes <a moz-do-not-send="true"
href="https://halldweb.jlab.org/wiki/index.php/HDGeant4_Meeting,_March_26,_2019#Minutes">here</a>
        and below.</p>
      <p>  -- Mark</p>
      <p>_____________________________________</p>
      <p> </p>
      <div id="globalWrapper">
        <div id="column-content">
          <div id="content" class="mw-body" role="main">
            <h1 id="firstHeading" class="firstHeading" lang="en"><span
                dir="auto">Minutes, HDGeant4 Meeting, March 26, 2019</span></h1>
            <div id="bodyContent" class="mw-body-content">
              <div id="mw-content-text" dir="ltr" class="mw-content-ltr"
                lang="en">
                <p>Present: </p>
                <ul>
                  <li> <b> JLab: </b> Alex Austregesilo, Thomas
                    Britton, Sean Dobbs, Mark Ito (chair),Richard Jones,
                    Simon Taylor</li>
                </ul>
                <p>There is a <a rel="nofollow" class="external text"
                    href="https://bluejeans.com/s/96_2v"
                    moz-do-not-send="true">recording of this meeting</a>
                  on the BlueJeans site. Use your JLab credentials to
                  access it. </p>
                <h3><span class="mw-headline"
                    id="Review_of_minutes_from_the_last_meeting">Review
                    of minutes from the last meeting</span></h3>
                <p>We went over <a
href="https://halldweb.jlab.org/wiki/index.php/HDGeant4_Meeting,_March_12,_2019#Minutes"
                    title="HDGeant4 Meeting, March 12, 2019"
                    moz-do-not-send="true">the minutes</a> without
                  significant comment. </p>
                <h3><span class="mw-headline"
                    id="Do-over_on_comparison_studies">Do-over on
                    comparison studies</span></h3>
                <p>Simon has run some single-particle gun simulations
                  for both HDG3 and HDG4 on the OSG. He has not had a
                  chance to look at them in detail yet. His initial
                  impression is that differences are seen in
                  calorimetry. </p>
                <h3><span class="mw-headline"
                    id="Further_comparative_studies_of_calorimeter_response">Further
                    comparative studies of calorimeter response</span></h3>
                <p>Richard showed <a rel="nofollow" class="external
                    text"
href="https://docs.google.com/presentation/d/1IDw4LD2GhBpzam9E1V8LRBcb49Kb1V3GMTWR2Inw2V8/edit?usp=sharing"
                    moz-do-not-send="true">some slides</a> describing
                  extensions of his work on low-level comparisons of
                  HDG3 and HDG4 for calorimetry. </p>
                <ul>
                  <li> He looked at the residual non-linearity, albeit
                    up to 8 GeV, for reconstructed shower energy in the
                    FCAL. Agreement between HDG3 and HDG4 is very good
                    now, at all energies.</li>
                  <li> He has examined cluster time, relative to "RF".
                    Both HDG3 and 4 show the same linear systematic
                    dependence in relative time as a function of energy.</li>
                  <li> For BCAL energy response, he showed the same
                    residual non-linearity study as shown earlier for
                    the FCAL. Here the qualitative agreement is good,
                    with perhaps an 8 MeV upward shift of HDG4 relative
                    to HDG3. The effect of energy leakage out the back
                    of BCAL dominates the changes in the trends as polar
                    angle moves from far forward to 90 degrees.</li>
                  <li> For BCAL timing, he noted a difference in the
                    algorithm used to get cluster timing between HDG3
                    and HDG4
                    <ul>
                      <li> HDG3 uses energy-weighted average shower
                        times</li>
                      <li> HDG4 uses the earliest hit per sector for
                        times</li>
                    </ul>
                  </li>
                  <li> We noted that mcsmear smears the BCAL timing by
                    55 ps, not a lot. Richard sees this in his study.</li>
                  <li> The difference in the BCAL algorithms does not
                    make much apparent difference. He suggests we use
                    energy weighted for both to be consistent.</li>
                  <li> Richard is planning to deprecate the
                    bcalSiPMUpHit and bcalSiPMDownHit elements in HDDM
                    in favor of a single bcalTruthCell element that can
                    serve information from both upsteam and downstream
                    ends of the BCAL,</li>
                  <li> There is still an issue that Alex sees with the
                    ρ's albeit with charged tracks and not
                    electromagnetic showers. Hopefully Simon's
                    single-track studies will address it.</li>
                </ul>
              </div>
              <div class="printfooter"> Retrieved from "<a dir="ltr"
href="https://halldweb.jlab.org/wiki/index.php?title=HDGeant4_Meeting,_March_26,_2019&oldid=92045"
                  moz-do-not-send="true">https://halldweb.jlab.org/wiki/index.php?title=HDGeant4_Meeting,_March_26,_2019&oldid=92045</a>"</div>
            </div>
          </div>
        </div>
        <div id="footer" role="contentinfo">
          <ul id="f-list">
            <li id="lastmod"> This page was last modified on 29 March
              2019, at 13:25.</li>
          </ul>
        </div>
      </div>
    </blockquote>
  </body>
</html>