[Paraview] Sorry for the 2 previous posts (this is the one)

[Nick Franklin]

Sorry for the two previous posts I am new to the forum and it took me 2
tries before I realized I needed to turn on plain text.

Paraview - render-server/data-server

I have a few questions regarding Paraview and the way it works in
client/server mode.

My overall goal is to setup Paraview to distribute the data calculations
(e.g. filters, manipulation of large data sets, etc) and the rendering of
our data across a cluster of linux boxes.

I will start with what I have done:  I have configured Paraview with MPI for
linux (RH 9.1) and Windows.  I have configured a setup wherebye I run
Paraview in server mode and it distributes to 6 nodes.  Here is the command
I am using:  mpirun -np 6 -p4ssport xxxx paraview --server

The -p4ssport bit is a way of bypassing potential ssh problems that could
come up.  It simply uses mpich's secure server to run shell commands on
other servers, the only dissadvantate (I think) is that I have to manually
key in the passwords for each node everytime I start a Paraview server.

So.the server starts.I type in the password for each node and then it says
"waiting for client on port 11111" and instances of paraview popup in the
system monitor on each linux node.

I then connect to the server using the command: paraview --client  the
client connects successfully and all seems well and good.  I make a
ridiculously large sphere on the client and the system monitor on all the
nodes start to wizz up to 100%.  Then 6 tiles appear on the server machine
containing 6 slices of sphere.  Then the final sphere appears on the client
machine.

So.what is happening here???  Is the rendering of the sphere being
distributed or is the data generation of the sphere being distributed.

Does paraview supportdistributed data manipulation (eg, If I wanted to
extract edges on a very high polygon shape would it distribute the
calculation of doing this).  Or does it only allow distributed rendering of
the data once it has been manipulated on a single machine?

I have tried to set up a data and render server and then connect using the
-crs argument and it does work but only the nodes of the render server do
anything and the data-server nodes (the nodes started with the command
--server) don't appear to be doing anything.  But the rendernodes started by
the -render-server command seem to at least have some activity.

Could anyone please shed light on how exactly Paraview works in this way.

Is there some classic way of doing what I am trying todo?  How are others
doing this?

Also if anyone is interested in instructions on compiling Paraview with MPI
on RH Linux 9.1 let me know because it seems that there might be demand for
this looking through the discussion groups.

The setup I am using is a 6 node (to begin with), Athlon 64 3000 each with
0.5gb of ram and all running RH 9.1.  As the client machine I am using a
windows Pentium 4 system.  The version of Paraview is 1.6.1 and it is
compiled with MPICH 1.x.x.

Any help would be great.

Regards

Nick

---------------------------------

Nick Franklin

Email:  nickfzx at aol.com <mailto:nickfzx at aol.com> 

Tel: 415 640 3124

---------------------------------