<div dir="ltr">Thanks for the reference. What exactly is the benefit of having the data in appended mode? I guess right now I'm using  binary mode</div><div class="gmail_extra"><br><br><div class="gmail_quote">On Mon, May 5, 2014 at 11:46 AM, Burlen Loring <span dir="ltr"><<a href="mailto:bloring@lbl.gov" target="_blank">bloring@lbl.gov</a>></span> wrote:<br>

<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
  
    
  
  <div text="#000000" bgcolor="#FFFFFF">
    Ahh, right, I assumed you were using VTK's classes to write the
    data, although if you're writing them yourself, you'll still want to
    emulate the "unencoded appended" format to get the best performance.
    see SetModeTo* and EncodeAppendedData* funcs (<a href="http://www.vtk.org/doc/nightly/html/classvtkXMLWriter.html" target="_blank">http://www.vtk.org/doc/nightly/html/classvtkXMLWriter.html</a>)<br>
    <br>
    There's also some essential info in the file format doc(
    
    <a href="http://www.vtk.org/VTK/img/file-formats.pdf" target="_blank">http://www.vtk.org/VTK/img/file-formats.pdf</a>).
    Search for "Appended".<br>
    <br>
    One way to get a handle on what's happening with the various modes
    and options is to examine the files you can produce in PV. For
    example open up PV, create a sphere source(or if you prefer some
    unstructured data), under file menu save the data and chose one of
    the pvt* options Compare the files produced for binary and appended
    modes etc...<div><div class="h5"><br>
    <br>
    <div>On 05/05/2014 11:34 AM, Mohammad
      Mirzadeh wrote:<br>
    </div>
    <blockquote type="cite">
      <div dir="ltr">
        <div class="gmail_extra">
          <div class="gmail_quote">
            <div>Burlen,</div>
            <div><br>
            </div>
            <div>Thanks a lot for your comments. </div>
            <div> </div>
            <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
              <div text="#000000" bgcolor="#FFFFFF"> This is not an
                answer to your question but there is a usage caveat w/
                VTK XML files that I want to make sure you're aware of. 
                When you use that format make sure you set mode to
                "appended" and "encode" off. This is the combination to
                produce binary files which are going to be faster and
                very likely smaller too. You probably already know that,
                but just in case ...<br>
                <br>
              </div>
            </blockquote>
            <div> </div>
            <div>I write the data itself as binary inside the .vtu file
              itself. Is this what you mean by appended mode? I cannot
              see any 'mode' keyword the xml file. Same as for encode; I
              don't have it in the xml file.</div>
            <div><br>
            </div>
            <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
              <div text="#000000" bgcolor="#FFFFFF"> now to get to your
                question:
                <div><br>
                  <blockquote type="cite">1) Go with a single parallel
                    HDF5 file that includes data for all time-steps.
                    This makes it all nice and portable except there are
                    two issues. i) It looks like doing MPI-IO might not
                    be as efficient as separate POSIX IO, especially on
                    large number of processors. ii) ParaView does not
                    seem to be able to read HDF5 files in parallel</blockquote>
                </div>
                comment: If I were you I'd avoid putting all time steps
                in a single file, or any solution where files get too
                big. Once files occupy more than ~80% of a tape drive
                you'll have very hard time getting them on and off
                archive systems. see this: <a href="http://www.nersc.gov/users/data-and-file-systems/hpss/storing-and-retrieving-data/mistakes-to-avoid/" target="_blank">http://www.nersc.gov/users/data-and-file-systems/hpss/storing-and-retrieving-data/mistakes-to-avoid/</a>
                My comment assumes that you actually use such systems.
                But you probably will need to if you generate large
                datasets at common HPC centers.<br>
                <br>
              </div>
            </blockquote>
            <div><br>
            </div>
            <div>That's actually a very good point I was not thinking
              of! Thanks for sharing.</div>
            <div> </div>
            <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
              <div text="#000000" bgcolor="#FFFFFF"> I've seen some AMR
                codes get elaborate in their HDF5 formats and run into
                serious performance issues as a result. So my comment
                here is that if you go with HDF5, keep the format as
                simple as possible! and of course file sizes small
                enough to be archived ;-)<span><font color="#888888"><br>
                    <br>
                    Burlen</font></span>
                <div>
                  <div><br>
                    <br>
                    <br>
                    <div>On 05/05/2014 10:48 AM, Mohammad Mirzadeh
                      wrote:<br>
                    </div>
                    <blockquote type="cite">
                      <div dir="ltr">They are represented as
                        unstructured grid. As a sample run, a 100M grid
                        point on 256 proc produces almost 8.5G file. We
                        intent to push the limits close to 1B at most at
                        this time with # processors up to a few
                        thousands. However, it would be good to have
                        something that could scale to larger problems as
                        well</div>
                      <div class="gmail_extra"><br>
                        <br>
                        <div class="gmail_quote">On Sat, May 3, 2014 at
                          1:28 AM, Stephen Wornom <span dir="ltr"><<a href="mailto:stephen.wornom@inria.fr" target="_blank">stephen.wornom@inria.fr</a>></span>
                          wrote:<br>
                          <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Mohammad
                            Mirzadeh wrote:<br>
                            <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
                              <div>
                                <div> Hi I am at a critical point in
                                  deciding I/O format for my
                                  application. So far my conclusion is
                                  to use parallel HDF5 for restart files
                                  as they are quite flexible and
                                  portable across systems.<br>
                                  <br>
                                  When it comes to visualization,
                                  however, i'm not quite sure. Up until
                                  now I've been using pvtu along with
                                  vtu files and although they generally
                                  work fine, one easily gets in trouble
                                  when running big simulations on large
                                  number of processors as the number of
                                  files can easily get out of control
                                  and even simplest utility commands
                                  (e.g. ls) takes minutes to finish!<br>
                                  <br>
                                  After many thinking I've come to a
                                  point to decide between two
                                  strategies:<br>
                                  <br>
                                  1) Go with a single parallel HDF5 file
                                  that includes data for all time-steps.
                                  This makes it all nice and portable
                                  except there are two issues. i) It
                                  looks like doing MPI-IO might not be
                                  as efficient as separate POSIX IO,
                                  especially on large number of
                                  processors. ii) ParaView does not seem
                                  to be able to read HDF5 files in
                                  parallel<br>
                                  <br>
                                  2) Go with the same pvtu+vtu strategy
                                  except take precautions to avoid file
                                  explosions. I can think of two
                                  strategies here: i) use nested folders
                                  to separate vtu files from pvtu and
                                  also each time step ii) create an IO
                                  group communicator with much less
                                  processors that do the actual IO.<br>
                                  <br>
                                  My questions are 1) Is the second
                                  approach necessarily more efficient
                                  than MPI-IO used in HDF5? and 2) Is
                                  there any plan to support parallel IO
                                  for HDF5 files in paraview?<br>
                                  <br>
                                  <br>
                                </div>
                              </div>
                              <div>
                                _______________________________________________<br>
                                Powered by <a href="http://www.kitware.com" target="_blank">www.kitware.com</a><br>
                                <br>
                                Visit other Kitware open-source projects
                                at <a href="http://www.kitware.com/opensource/opensource.html" target="_blank">http://www.kitware.com/opensource/opensource.html</a><br>
                                <br>
                                Please keep messages on-topic and check
                                the ParaView Wiki at: <a href="http://paraview.org/Wiki/ParaView" target="_blank">http://paraview.org/Wiki/ParaView</a><br>
                                <br>
                                Follow this link to
                                subscribe/unsubscribe:<br>
                                <a href="http://www.paraview.org/mailman/listinfo/paraview" target="_blank">http://www.paraview.org/mailman/listinfo/paraview</a><br>
                              </div>
                            </blockquote>
                            Are your meshes structured or unstructured?
                            How many vertices in your meshes?<span><font color="#888888"><br>
                                <br>
                                Stephen<br>
                                <br>
                                -- <br>
                                <a href="mailto:stephen.wornom@inria.fr" target="_blank">stephen.wornom@inria.fr</a><br>
                                2004 route des lucioles - BP93<br>
                                Sophia Antipolis<br>
                                06902 CEDEX<br>
                                                <br>
                                Tel: 04 92 38 50 54<br>
                                Fax: 04 97 15 53 51<br>
                                <br>
                              </font></span></blockquote>
                        </div>
                        <br>
                      </div>
                      <br>
                      <fieldset></fieldset>
                      <br>
                      <pre>_______________________________________________
Powered by <a href="http://www.kitware.com" target="_blank">www.kitware.com</a>

Visit other Kitware open-source projects at <a href="http://www.kitware.com/opensource/opensource.html" target="_blank">http://www.kitware.com/opensource/opensource.html</a>

Please keep messages on-topic and check the ParaView Wiki at: <a href="http://paraview.org/Wiki/ParaView" target="_blank">http://paraview.org/Wiki/ParaView</a>

Follow this link to subscribe/unsubscribe:
<a href="http://www.paraview.org/mailman/listinfo/paraview" target="_blank">http://www.paraview.org/mailman/listinfo/paraview</a>
</pre>
                    </blockquote>
                    <br>
                  </div>
                </div>
              </div>
            </blockquote>
          </div>
          <br>
        </div>
      </div>
      <br>
      <fieldset></fieldset>
      <br>
      <pre>_______________________________________________
Powered by <a href="http://www.kitware.com" target="_blank">www.kitware.com</a>

Visit other Kitware open-source projects at <a href="http://www.kitware.com/opensource/opensource.html" target="_blank">http://www.kitware.com/opensource/opensource.html</a>

Please keep messages on-topic and check the ParaView Wiki at: <a href="http://paraview.org/Wiki/ParaView" target="_blank">http://paraview.org/Wiki/ParaView</a>

Follow this link to subscribe/unsubscribe:
<a href="http://www.paraview.org/mailman/listinfo/paraview" target="_blank">http://www.paraview.org/mailman/listinfo/paraview</a>
</pre>
    </blockquote>
    <br>
  </div></div></div>

</blockquote></div><br></div>