[Paraview] Question about paraview's scalability.

[Raashid Baig]

I have finally compiled ParaView3.2.1 on a cluster with OSMesa and mpi support. It works without any X server problem on 8 nodes. 
I run paraview as 

/usr/local/openmpi-1.2.4/bin/mpirun -np 8 --x LD_LIBRARY_PATH=/home/raashid/src/lib64/:$LD_LIBRARY_PATH --mca btl ^openib,mvapi -hostfile nodes ./pvserver --use-offscreen-rendering

I have generated a vtk-xml structured grid file with about 48 million cells (file size about 800MB) which I am able to visualize without any problem when I run pvserver on 8 nodes. I have an ensight file with 33 million cells (Unstructured grid, generated by another software, but is a lot larger in terms of disk space usage about 2.1GB). ParaView fails  while trying to load this file. pvserver crashes when the memory usage on the headnode of the cluster is about 20% to total available memory. 

Are the only some file formats which can be read and visualized in parallel efficiently ? 

I have checked the instructions given in 
http://www.paraview.org/Wiki/Setting_up_a_ParaView_Server
about paraview's parallel rendering set-up, and everything appears to be working as suggested.  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/paraview/attachments/20080114/cbe0a288/attachment.html