[Bioclusters] mpiblast on the smallest possible cluster

Fernan Aguero bioclusters@bioinformatics.org
Fri, 23 Jan 2004 17:34:58 -0300


I'm trying mpiblast for the first time, and I'm testing it
in a small (hmm, indeed very small) cluster. Actually, you
wouldn't call it a cluster :)

I have two machines. A master node, and a slave node. I have
installed mpich-1.2.5 and mpiblast-1.2.0 on both. 

All the tests I've performed (from reading the docs) pass.
However, all the jobs run on the slave. The master just sits
down quietly waiting for the slave to finish with one
segment, before feeding the next. I suppose that this
is the desired behaviour when you have a big cluster --
you'll want to use the master node for managing the nodes,
and not run any job.

But what about small clusters of workstations? I would want
the jobs to be run also by the master host. Is this
impossible to do with the current state of the tools? I
imagine that for a small number of nodes, and if the master
is not managing anything else (sge or other mpi jobs), it
should be possible to run blast jobs on the master without
affecting its ability to manage the slaves. Am i right here?

Thanks in advance for any tip/suggestion,


F e r n a n   A g u e r o