[Paraview] ParaView Research Project

[Moreland, Kenneth]

Generally the best place to ask questions is the ParaView mailing list (paraview at paraview.org).

To quickly answer your question, the data partitioning is usually chosen by the reader.  The ParaView/VTK system simply asks the reader to load piece i of n, where i is generally set to the rank of the MPI process.

Things are different for structured grids (i.e. uniform, rectilinear, and curvilinear grids).  It's more complicated due to the way VTK handles these type of extents.

-Ken

From: Thomas Fronk <fronk at mit.edu<mailto:fronk at mit.edu>>
Date: Wednesday, November 7, 2012 11:11 AM
To: Kenneth Moreland <kmorel at sandia.gov<mailto:kmorel at sandia.gov>>, Kenneth Moreland <bnwylie at sandia.gov<mailto:bnwylie at sandia.gov>>, "dhrogers at sandia.gov<mailto:dhrogers at sandia.gov>" <dhrogers at sandia.gov<mailto:dhrogers at sandia.gov>>
Subject: [EXTERNAL] ParaView Research Project

Hello,

My name is Thomas Fronk and I am a senior at Massachusetts Institute of Technology (MIT). I am currently working on a research project concerning ParaView for the Institute for Soldier Nanotechnologies (ISN). I am unsure if I have the correct contact information with regards to ParaView advice/support and I apologize in advance for any inconveniences.

I have a question about the way ParaView partitions data on multiple nodes. It seems to partition the data automatically without the user having any control.

To give you some background, my research project is to determine whether it is possible to optimize processing time through pre-splitting up a large data set. We want to tell ParaView how to split up a large data set unto multiple nodes and then run ParaView in parallel (reverse client/server mode) telling it that it should use those specific nodes. The goal is to have the individual nodes perform their calculation and rendering  with the local data fragment that was previously stored there, thus avoiding additional delays caused by the nodes having to access the head node where the entire data is stored.

Is this possible? Is there a way to mandate which nodes to use and how ParaView should split up the data among the nodes? Also, can ParaView tell the individual nodes to work with the locally stored data fragment instead of accessing the whole data set at the head node?

Thank you very much for your help and I look forward to hearing from you.

Best Regards,
Thomas

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20121107/ad72d505/attachment.htm>