UCSF: Cluster for distributed biomedical computing

R - LAM/MPI tutorial by Steve Shiboski: Step-by-step instructions on how to distribute computations using R on our cluster
MPI homepage at Argonne National Laboratory: Message Passing Interface; library specs for "message passing"
MPICH and LAM-MPI: Two implementations of MPI.
Simple Parallel Statistical Computing in R by Rossini, Tierney, and Li: Technical report describing the "snow" package in R.
Simple Network of Workstations in R by Luke Tierney: Describes the setup at the University of Iowa

How big is is cluster?: We have six nodes; one master and five compute nodes.
I have some C code. Can I parallelize it?: Depends. If your program is such that parallel calls to a subroutine are made, then yes. One way would be to either parallelize it with a shell script, and collect the job results yourself. The second way would be to run the job from within R and let it do the collection work.
How fast is the cluster?: Each processor is approximately four times faster than a processor on our computing server, Sheep (which is a Sun Enterpise 420R server with four 450-MHz UltraSPARCII processors, 4 GB of RAM). This means that the cluster is about 12 times faster than all four Sheep processors combined. In our test runs, the gain has been somewhare between 10 to 12 times.
Do you have SAS?: Sorry, we do not have SAS on the cluster now, as it does not run on Mac platforms.
Do you have Matlab?: We don't. Matlab is available for Macs, but it would be quite expensive for us to install on our cluster. If you are willing to pay for the installation, we will do it for you!

--- documentation ---