nfs mounted executables

[uccatvm]

Hi all,

We are setting up a cluster of dual Xeon Linux boxes connected by 
fast ethernet. At the moment we only have 7 nodes, but we hope to
grow this to, say, 32 nodes. This cluster will be used by several 
research groups running a variety of different codes, both in-house and 
commercial packages (like Gaussian), some parallel, some not. Most 
calculations will be either sequential, or parallel within 1 node, 
but some calculations will be done over several nodes.  

Now, my question is, what is the best way of managing the programs.
Is it best to install all programs in a directory on the master that 
is nfs mounted to all slaves? I would think that this wouldn't affect 
performance, as programs are loaded into memory when a calculation starts,
and barely need any further network traffic. Is this correct?
The alternative would be to install all executables on all slave nodes. 
The disadvantage of this approach is that it takes away disk space on 
the slaves. 

I would be much interested in hearing your opinion on this.

Tanja
-- 
  =====================================================================
     Tanja van Mourik                                                
     Royal Society University Research Fellow
     Chemistry Department 
     University College London    phone:    +44 (0)20-7679-4663      
     20 Gordon Street             e-mail:   work: T.vanMourik at ucl.ac.uk 
     London WC1H 0AJ, UK                    home: tanja at netcomuk.co.uk     

     http://www.chem.ucl.ac.uk/people/vanmourik/index.html
  =====================================================================