<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Hi Felix,</p>
    <p>Thanks for your mail. I will test it more to make sure that it
      doesn't break anything. Also I have added a configuration key for
      easier switching to the older code in case if there is any
      problem. If you can help me in any manner in testing or
      performance numbers, please let me know.</p>
    <p><br>
    </p>
    <p>Regards</p>
    <p>Rafi KC<br>
    </p>
    <div class="moz-cite-prefix">On 27/05/20 12:37 pm, Felix Kölzow
      wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:d2d05f50-e2ee-4f99-7c51-471f69244e4b@gmx.de">
      <p>Dear Rafi,</p>
      <p><br>
      </p>
      <p>thanks for your effort. I think this is of great interest of
        many gluster users. Thus, I would really encourage you to <br>
      </p>
      <p>test and to further improve this feature. Maybe it is
        beneficial to create a certain guideline which things should be
        tested</p>
      <p>to make this feature really ready for productive use.</p>
      <p><br>
      </p>
      <p>Thanks in advance.</p>
      <p>Felix<br>
      </p>
      <div class="moz-cite-prefix">On 27/05/2020 07:56, RAFI KC wrote:<br>
      </div>
      <blockquote type="cite"
        cite="mid:9e1d43b8-98cc-9dae-193e-75faf6793e59@redhat.com">
        <p>Hi All,</p>
        <p>I have been working on POC to improve readdirp performance
          improvement. At the end of the experiment, The results are
          showing promising result in performance, overall there is a
          104% improvement for full filesystem crawl compared to the
          existing solution. Here is the short test numbers. The tests
          were carried out in 16*3 setup with 1.5 Million dentries (Both
          files and dir). The system also contains some empty
          directories. <b>In the result the proposed solution is 287%
            faster than the plane volume and 104% faster than the
            parallel-readdir based solution.</b><br>
        </p>
        <p><b
            id="docs-internal-guid-8644ede0-7fff-09be-a0d9-5ee943340f2c"><br>
          </b></p>
        <div dir="ltr">
          <table>
            <colgroup><col width="199"><col width="113"><col><col></colgroup><tbody>
              <tr>
                <td>
                  <p dir="ltr"><span>Configuration</span></p>
                </td>
                <td>
                  <p dir="ltr"><span>Plain volume</span></p>
                </td>
                <td>
                  <p dir="ltr"><span>Parallel-readdir</span></p>
                </td>
                <td>
                  <p dir="ltr"><span>Proposed Solution</span></p>
                </td>
              </tr>
              <tr>
                <td>
                  <p dir="ltr"><span>FS Crawl Time in Seconds</span></p>
                </td>
                <td>
                  <p dir="ltr"><span>16497.523</span></p>
                </td>
                <td>
                  <p dir="ltr"><span>8717.872</span></p>
                </td>
                <td>
                  <p dir="ltr"><span>4261.401 </span></p>
                </td>
              </tr>
            </tbody>
          </table>
        </div>
        <p>In short, the basic idea behind the proposal is the efficient
          managing of readdir buffer in gluster along with prefetching
          the dentries for intelligent switch-over to the next buffer.
          The detailed problem description, deign description and
          results are available in the doc.<a
href="https://docs.google.com/document/d/10z4T5Sd_-wCFrmDrzyQtlWOGLang1_g17wO8VUxSiJ8/edit"
            moz-do-not-send="true">https://docs.google.com/document/d/10z4T5Sd_-wCFrmDrzyQtlWOGLang1_g17wO8VUxSiJ8/edit</a>
          <br>
        </p>
        <p><br>
        </p>
        <p>If anybody can help with the testing on a different kind of
          workloads, I would be very happy to assist. If wanted to test
          the patch and run a performance test on your setup, I could
          help with back-porting the patch to the version of your
          choice.<br>
        </p>
        <p><br>
        </p>
        <p><a class="moz-txt-link-freetext"
            href="https://review.gluster.org/24469"
            moz-do-not-send="true">https://review.gluster.org/24469</a></p>
        <p><a class="moz-txt-link-freetext"
            href="https://review.gluster.org/24470"
            moz-do-not-send="true">https://review.gluster.org/24470</a><br>
        </p>
        <p><br>
        </p>
        <p>Regards</p>
        <p>Rafi KC<br>
        </p>
        <br>
        <fieldset class="mimeAttachmentHeader"></fieldset>
        <pre class="moz-quote-pre" wrap="">________



Community Meeting Calendar:

Schedule -
Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC
Bridge: <a class="moz-txt-link-freetext" href="https://bluejeans.com/441850968" moz-do-not-send="true">https://bluejeans.com/441850968</a>

Gluster-users mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Gluster-users@gluster.org" moz-do-not-send="true">Gluster-users@gluster.org</a>
<a class="moz-txt-link-freetext" href="https://lists.gluster.org/mailman/listinfo/gluster-users" moz-do-not-send="true">https://lists.gluster.org/mailman/listinfo/gluster-users</a>
</pre>
      </blockquote>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <pre class="moz-quote-pre" wrap="">________



Community Meeting Calendar:

Schedule -
Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC
Bridge: <a class="moz-txt-link-freetext" href="https://bluejeans.com/441850968">https://bluejeans.com/441850968</a>

Gluster-users mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Gluster-users@gluster.org">Gluster-users@gluster.org</a>
<a class="moz-txt-link-freetext" href="https://lists.gluster.org/mailman/listinfo/gluster-users">https://lists.gluster.org/mailman/listinfo/gluster-users</a>
</pre>
    </blockquote>
  </body>
</html>