[lug] [CLUE-Tech] Super Computing Event: SC2001 (fwd)

[Sexton, George]

I actually have experience configuring MPI and PVM on clusters. I'm also in
the process of creating an RPM for the HPL Linpack program for RedHat 7.x.
After I finish that, I am going to create an RPM of MPI/POVRAY. Both would
be handy for demo/testing of a cluster.

I can also give tips on how to configure the system.

Anyone with two computers (or a dual processor box) can create a "Cluster".
The basic idea is:

install MPI libraries on nodes.
Enable remote login on nodes though RSH or SSH.
Configure MPI
run your parallel program using the MPIRUN command. Note that the program
has to be available to every node. I.E. exported via NFS.

In a nutshell, that's it. Getting a basic cluster up and running is pretty
easy to do.

MPICH comes with demonstration programs that display the basics of using MPI
to create parallel programs. There are also some books in print on using MPI
to create programs.

>>I think Wayde mentioned trying to get a test cluster up for the BLUG to
>>use. Maybe a "Stone Soupercomputer" style cluster would be a good
>>learning aid. If BLUG can't do it alone, maybe in conjunction with the

-----Original Message-----
From: lug-admin at lug.boulder.co.us [mailto:lug-admin at lug.boulder.co.us]On
Behalf Of Charles Morrison
Sent: 13 November, 2001 11:07 AM
To: lug at lug.boulder.co.us
Subject: Re: [lug] [CLUE-Tech] Super Computing Event: SC2001 (fwd)


Unfortunately, this is how it always is; You can't get the skills unless
you work in the area and you can't work in the area unless you already
have the skills. It's really a PITA.

The main emphasis for this opening is PBS and other things that are
quite specific to distributed processing environments, ie Beowulf
clusters. BTW, these kind of clusters are being used more and more
outside of gov't labs nowdays. They're used in genetic research, oil
companies, financial firms, moviemaking and virtually anywhere there is
a need for big computing resources and a desire to save money. While the
job market for these systems isn't huge, it is there.

I think Wayde mentioned trying to get a test cluster up for the BLUG to
use. Maybe a "Stone Soupercomputer" style cluster would be a good
learning aid. If BLUG can't do it alone, maybe in conjunction with the
other LUGS and other special interest UGs (ACM) as well. Might be worth
a discussion again.