[Bioclusters] Clusters for bioinformatics... Some numbers or statistics?

Ivo Grosse grosse@cshl.org
Thu, 30 Aug 2001 21:19:14 -0400

Hi Jim, Jeremy, and Chris,

thanks for your great and detailed answers.  As always, they create 
even more questions.  :-)  Before asking them, let me clarify that we 
are not thinking of running jobs on all of-the-order-of 1000 machines 
of our lab, but only on our of-the-order-of 10 machines of our group.

> the try, and Condor sounds like a good approach.  The cost of the new
> computers (easy to define) vs. risk to your clients data, processes, and
> systems (harder to define) has to be balanced.

I still would like to understand more clearly the risk.  We are not 
talking about O(1000) machines scattered around in dozens of buildings 
and used by people who we don't know and who don't know us, but we are 
talking about our own group.  At BU we used the same concept, with 50 - 
100 machines, and it worked quite well, i.e., I don't remember any 
intereference of the back ground with the foreground jobs or vice 
versa, so I would like to learn more about potential risks.  Here, our 
group is much smaller, so I think the risk of someone rebooting the 
machines is pretty small.

Which other disadvantages does a network of workstations have over a