<br><div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div><p>1) I want to make Paraview run in render-server mode and have N machines (i.e
., apart from the data server and the client) to perform rendering. The Paraview Guide says that there are two connections must be make for Paraview to run in render-server mode. The first connection is between the client and the first node of each of the data and render servers. The second connection is between the nodes of the render server and "the first N nodes" of the data server.
</p><p>Supposed that there is 5 data servers, DS0~DS4, and 2 render servers, RS0 and RS1. N=2, M=5. </p><p> </p><p>I think the DS0 sends its data to RS0 for rendering, and DS1 sends to RS1. right? then How do the other data servers ( DS3,DS4 and DS5 ) distribute their data?
</p></div></blockquote><div><br>The geometry produced by DS0-DS4 is redistributed to DS0 and DS1. It is then sent to the render server (DS0 -> RS0, DS1 -> RS1).<br></div><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div><p>2) when paraview run in render-server mode, it requires that the number (N) of render servers must be no more than the number (M) of data servers. Now I want to use a SGI Altix 4700 as the data server. It is a SMP machine, has 16 CPUs but no graphic cards. And I use a 4-nodes cluster equipped with 4 graphic cards as the render servers.
</p><p>Now N=4 and M=1. </p><p> </p><p>how to make paraview run in render-server mode on there machines? </p><p> </p></div></blockquote></div>You can still run MPI on an SMP machine. Make sure that the MPI distribution makes use of shared memory to improve performance. You have to do this in order to make use of multiple CPUs anyway. ParaView disables all multi-threading support that comes with VTK.
<br><br>-- <br> Berk Geveci<br> Kitware Inc.<br> 28 Corporate Drive<br> Clifton Park, NY, 12065