non-standard cluster configurations

[Jason Abate]

> We're looking to configure two clusters of approx 30 machines each cluster.
> What we would like to do is have each cluster be able to operate
> independantly (easiest scenario); or to have all 60 machines operate as a
> single cluster with one master server; and then also take each cluster and
> treate one as a slave and one as the server (have the entire 30 node cluster
> act as a server to the other cluster.)

Off the top of my head, I can't see why you'd want to setup things in 
this manner (I'm sure you have a good reason - I just can't see it!). 
If this is to limit the number of processors available to different 
users (i.e. part of the time you can run up to 30-node jobs, at other 
times you can run 60-node jobs), it seems like you would be better 
suited to setting up the system as one big cluster and using a 
queueing system such as PBS or NQS with multiple queues supporting 
the different configurations.  

This way you only have one configuration to setup and support, yet you
still have the flexability of partitioning the nodes however you like.
If there are other reasons for the different setups, definately post
the details (especially the slave/server setup).  I personally haven't
heard of anyone with this type of setup, but there is quite a wide
range of uses and setups for clusters.

-jason

====================================================================
   Jason Abate abate@ticam.utexas.edu www.ticam.utexas.edu/~abate
     Texas Institute for Computational and Applied Mathematics
     304 SHC, University of Texas at Austin,  Austin, TX 78712
     Work: 512-471-6947  Home: 512-912-1012  Fax: 512-471-8694