[MITgcm-support] Memory bandwidth monitoring (was: Scalability on a new Sgi node)
Jason Goodman
goodman_jason at wheatonma.edu
Tue Jun 14 12:54:05 EDT 2011
The discussion of scalability brings up an interesting question: are
there any tools or utilities on common Linux systems which allow you to
monitor memory bus congestion? This seems to be the typical bottleneck
for MITGCM. I can do the usual scaling analysis, but getting direct
info straight from the motherboards would be useful.
I'm using Scyld Clusterware on CentOS 4.9.
More information about the MITgcm-support
mailing list