Surjit-
NAMD should run w/o problems on an OSCAR cluster. However, NAMD performance largely relies on the hardware you're using. You can run NAMD over fast ethernet on clusters of 32-64 nodes. If you want to scale further than that, it is best to have a high speed interconnect like Myrinet.
OSCAR doesn't currently have built in support for Myrinet, but you could certainly add it manually to an OSCAR cluster. I don't have a performance comparison with Scyld, but you can probably find more information on it at
http://www.ks.uiuc.edu/Research/namd/tutorial/NCSA2001/
And here are some slides explicitly talking about designing a cluster for NAMD use:
http://www.ks.uiuc.edu/Research/namd/tutorial/NCSA2001/clusters.pdf
I couldn't find any benchmarks for NAMD on fast ethernet clusters... just Myrinet. The NCSA pt cluster is made up of dual PIII 1 Ghz machines, to give you a reference. It is a Beowulf I type cluster, similar to OSCAR. (except the Pt cluster has Myrinet)
http://www.ks.uiuc.edu/Research/namd/performance.html
Hope that helps-
Jeremy
At 01:42 PM 3/22/2002 -0500, Surjit Dixit wrote:
Hi,
I am working on setting up a PC cluster for performing parallel molecular dynamics simulations using the Amber or NAMD programs compiled with MPI.
I would like to know if anybody on this list has had any experience and success executing this application on Oscar. I would also appreciate any comments on the performance difference between an Oscar cluster and Scyld cluster for the same application. Would these benchmark alike or would they be different?
Thank you for your time.
Surjit
_______________________________________________
Oscar-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/oscar-users
_______________________________________________ Oscar-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/oscar-users
</x-flowed>
