From: Alessandro Cembran (cembran_at_chem.umn.edu)
Date: Wed Sep 20 2006 - 11:03:53 CDT
Thank you for the note. Thanks to Mark advices and then talking with my
sysadmin again, we figured out that the problem is due to the fact that
the 256 procs altix node here is not making use of the CPUset
functionality. Using that would most likely solve this problem, as also
shown by this pdf of the University of Innsbruck I found online:
Thanks everybody for the support, I will let you know how things are
going when (hopefully soon) the CPUset is implemented here.
Jim Phillips wrote:
> The shared processors sounds like the problem.
> NCSA's Altix is set up to assign dedicated processors from the
> queueing system and we get consistent performance. Their user docs
> are at http://www.ncsa.uiuc.edu/UserInfo/Resources/Hardware/SGIAltix/
> and you can probably have your sysadmin send them an email.
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:36 CST