:> Our lab is strongly considering purchasing a cluster of ~12 PC with dual PII
-: 450 MHz processors and 256 MB RAM each to configure in a Beowulf cluster run
-:ning Linux. Does anyone have experience running this type of system? Readin
-:g over past postings, I have come across people mentioning "possible" configu
-:rations, but I have never read where anyone has actually successfully ran CNS
-: over a cluster. I know that some software, such as Shelx, is made to run on
-: only one CPU. What about CNS? I would be very interested in hearing either
-: positive or negative comments from people who have experience running CNS on
-: a Beowulf cluster.
Depending on what you're interested in, you could use CHARMM instead
of CNS. Since I installed Josip Loncharic's TCP patches, I've seen
good scaling to 8 processors using dynamics with cutoffs, and good
scaling to 4 processors using Ewald sums. As I understand
it, CHARMM isn't nearly as easy to use for refinement, but the core
functionality (simulated annealing and NMR restraints) is present.
Alan Grossfield
--------------------------------------------------------------------------
|"In theory, there is no difference between theory and practice. In |
|practice, there is." Jan L.A. van de Snepscheut |
--------------------------------------------------------------------------
---
