<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div class="moz-cite-prefix">Ryan, <br>
      <br>
      Thanks for that tid-bit. I never thought of that. <br>
      <pre class="moz-signature" cols="72">
</pre>
      On 12/23/2014 09:14 PM, Novosielski, Ryan wrote:<br>
    </div>
    <blockquote
      cite="mid:B6DC3CAD-3E62-4E60-89D0-BEF4208CB7E8@ca.rutgers.edu"
      type="cite">
      <meta http-equiv="content-type" content="text/html; charset=utf-8">
      <div>I run an old Lustre (1.8.9), but it doesn't support some
        forms of file locking that were even required for compiling some
        software. Doesn't happen often, but enough to give me pause. <br>
        <br>
        <span style="background-color: rgba(255, 255, 255, 0);">____
          *Note: UMDNJ is now Rutgers-Biomedical and Health Sciences*<br>
          || \\UTGERS
               |---------------------*O*---------------------<br>
          ||_// Biomedical | Ryan Novosielski - Senior Technologist<br>
          || \\ and Health | <a moz-do-not-send="true"
            href="mailto:novosirj@rutgers.edu"
            x-apple-data-detectors="true"
            x-apple-data-detectors-type="link"
            x-apple-data-detectors-result="3">novosirj@rutgers.edu</a>-
          973/972.0922 (2x0922)<br>
          ||  \\  Sciences | OIRT/High Perf & Res Comp - MSB C630,
          Newark<br>
              `'</span></div>
      <div><br>
        On Dec 23, 2014, at 12:11, Prentice Bisbal <<a
          moz-do-not-send="true"
          href="mailto:prentice.bisbal@rutgers.edu">prentice.bisbal@rutgers.edu</a>>
        wrote:<br>
        <br>
      </div>
      <blockquote type="cite">
        <div><span>Beowulfers,</span><br>
          <span></span><br>
          <span>I have limited experience managing parallel filesytems
            like GPFS or </span><br>
          <span>Lustre. I was discussing putting /home and /usr/local
            for my cluster on </span><br>
          <span>a GPFS or Lustre filesystem, in addition to using it
            just for /scratch. </span><br>
          <span>I've never done this before, but it doesn't seem like
            all that bad an </span><br>
          <span>idea. My logic for this is the following:</span><br>
          <span></span><br>
          <span>1. Users often try to run programs from in /home, which
            leads to errors, </span><br>
          <span>no matter how many times I tell them not to do that.
            This would make the </span><br>
          <span>system more user-friendly. I could use quotas/policies
            to encourage them </span><br>
          <span>to use 'steer' them to use other filesystems if needed.</span><br>
          <span></span><br>
          <span>2. Having one storage system to manage is much better
            than 3.</span><br>
          <span></span><br>
          <span>3. Profit?</span><br>
          <span></span><br>
          <span>Anyway, another person in the conversation felt that
            this would be bad, </span><br>
          <span>because if someone was running a job that would hammer
            the fileystem, it </span><br>
          <span>would make the filesystem unresponsive, and keep other
            people from </span><br>
          <span>logging in and doing work. I'm not buying this concern
            for the following </span><br>
          <span>reasons:</span><br>
          <span></span><br>
          <span>If a job can hammer your parallel filesystem so that the
            login nodes </span><br>
          <span>become unresponsive, you've got bigger problems, because
            that means </span><br>
          <span>other jobs can't run on the cluster, and the job hitting
            the filesystem </span><br>
          <span>hard has probably slowed down to a crawl, too.</span><br>
          <span></span><br>
          <span>I know there are some concerns  with the stability of
            parallel </span><br>
          <span>filesystems, so if someone wants to comment on the
            dangers of that, too, </span><br>
          <span>I'm all ears. I think that the relative instability of
            parallel </span><br>
          <span>filesystems compared to NFS would be the biggest
            concern, not performance.</span><br>
          <span></span><br>
          <span>-- </span><br>
          <span>Prentice Bisbal</span><br>
          <span>Manager of Information Technology</span><br>
          <span>Rutgers Discovery Informatics Institute (RDI2)</span><br>
          <span>Rutgers University</span><br>
          <span><a moz-do-not-send="true" href="http://rdi2.rutgers.edu">http://rdi2.rutgers.edu</a></span><br>
          <span></span><br>
          <span>_______________________________________________</span><br>
          <span>Beowulf mailing list, <a moz-do-not-send="true"
              href="mailto:Beowulf@beowulf.org">Beowulf@beowulf.org</a>
            sponsored by Penguin Computing</span><br>
          <span>To change your subscription (digest mode or unsubscribe)
            visit <a moz-do-not-send="true"
              href="http://www.beowulf.org/mailman/listinfo/beowulf">http://www.beowulf.org/mailman/listinfo/beowulf</a></span><br>
        </div>
      </blockquote>
    </blockquote>
    <br>
  </body>
</html>