From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Mon Oct 15 2007 - 14:07:49 CDT
Hi Jerome,
I recently observed something similar running on abe. Jim's suspicion is
that it is something with the load balancer doing badly over long
periods. We'll be investigating it more thoroughly over the next couple
days. I'll give you a shout when a fix is ready.
Best,
Peter
Jerome Henin wrote:
> Hi all,
>
> Running NAMD on an NCSA cluster, I have noticed a significant drift in
> walltime/step within each run:
> http://www.cmm.upenn.edu/~jhenin/Files/NAMD-performance.png
>
> I wonder if anyone has an explanation for this.
>
> Thanks,
> Jerome
>
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:23 CST
