Intel and Myrinet cluster...

[Robert G. Brown]

Dear List Persons,

I am trying to help a group in our department and the chemistry
department make some design decisions concerning whether or not to go
with a beowulf-type architecture or an e.g. SP3.  Naturally, I
personally favor the former, but they are properly concerned about the
performance of their applications and want to test them on a cluster
architected according to one of their candidate designs.

At the moment they are considering a cluster architecture of Intel (or
possibly AMD) and Myrinet, to run some quantum chemistry codes.  The
codes themselves are apparently based on Gaussian 98 in one case and are
custom code written by one of the PI's in the other.  I know very little
about the internal architecture or the applications (language is C, I
believe, but I don't know if they are e.g. MPI applications yet and so
forth). I'm trying to find out.

At the moment I'm trying to find a test cluster with a similar
architecture that would loan them access long enough to generate some
comparative benchmarks (or compile a make-ready tarball of their
application and do a test run for them).  If anyone has such a cluster
and a bit of idle time and would like to volunteer it, I'd greatly
appreciate it.  Turnkey vendors are welcome -- if their alternative is
an SP3 I'd imagine that they would be perfectly happy to pay for a good
systems integration job that cost a third as much for equivalent
performance.

Please reply to me directly, not to the list.

    rgb

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu