[Paraview] Parallel File Formats in Paraview/VTK

[Berk Geveci]

Hi Luis,

The answer depends on various factors - data size, file system, number of
MPI ranks, do you need support for time series, do you mind dealing with
large number of files, does the topology change over time etc etc. Usually,
the optimal solution is to implement readers most suitable for these
factors. A custom HDF5 implementation for example.

However, implementing a new reader is a decent amount of work. At this
point, I suspect that the next best thing will be using the pvd format
pointing to vtm files point to vtu/vti/vtr/vts files. To avoid the file
explosion, you may want to reduce the number of vtu/vti/... files you write
by gathering data with MPI to a smaller MPI group before writing.
Unfortunately, there is currently no support for static topology so you
will have to duplicate that for each time step. If this is a major issue, I
recommend using the Exodus format for unstructured grids. It supports time
dependent and parallel data really well and is probably the best supported
format in ParaView.

We are currently working on a new set of readers and writers that will
support distributed data much better and that will be much more flexible to
address various needs. It will take us a few more months to get there
however.

Best,
-berk


On Wed, Dec 4, 2013 at 4:48 AM, Luis Manuel de Vries <
luis.de.vries at amphos21.com> wrote:

> Hi,
>
> I know that Paraview supports many different file formats (
> http://paraview.org/Wiki/ParaView:FAQ#What_file_formats_does_ParaView_support.3F
> ).
> I know that the following parallel binary formats are among the ones
> supported:
> - VTK files (xml with zipped, binary data)
> - ParaView data files
> - HDF5 files (different formats)
> - XDMF
>
> What I would like to know is which format is best parallelized by
> ParaView/VTK?
> When using structured or unstructured grids (size in the range of tens of
> gigabytes), does it
> make sense to split up the file in different data sets manually? Or does
> ParaView/VTK do this
> automatically? Or does it depend on which of the above formats is used?
> I have been using the PFLOTRAN HDF5 reader for a project (around 100
> million cells), and it
> did not behave very nicely so I was wondering if it would make sense in
> converting the files.
> Is there any difference in using XDMF (HDF5 + XML meta descriptions
> if I understood correctly) or HDF5 directly within ParaView?
>
> Thanks for the info,
>
> Luis Manuel
> ---------------------------------
> Luis Manuel de Vries
>
> AMPHOS21
> Passeig de Garcia i Faria, 49-51, 1-1
> 08019 Barcelona
> Spain
>
>
> Tel +34 935 830 500
> Fax +34 933 075 928
>
>
> *Luis.de.Vries at amphos21.com <Luis.de.Vries at amphos21.com>*
> www.amphos21.com
>
> [image: cid:image001.png at 01CA42C1.315780E0]
> http://www.amphos21.com/disclaimer.php
>
> _______________________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the ParaView Wiki at:
> http://paraview.org/Wiki/ParaView
>
> Follow this link to subscribe/unsubscribe:
> http://www.paraview.org/mailman/listinfo/paraview
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20131207/8c9c423c/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/png
Size: 9026 bytes
Desc: not available
URL: <http://www.paraview.org/pipermail/paraview/attachments/20131207/8c9c423c/attachment.png>