On Fri, 27 Feb 2009, Vo, Trinh wrote: VT> Dear Axel, VT> VT> Thanks for clarification. VT>
VT> About the benchmarks, I just simply to see how well is the VT> performance of the cluster we bought in term of scaling with QE. I VT> sent some plots to you, but the email did not go thru because of the VT> restriction of the size (larger than 40K). please send it to me personally. VT> Currently, I am not happy at the fact that the difference in CPU VT> time and wall time is too large. When I run a longer job, which what kind of hardware do you have, how much memory, and what kind of interconnect do you have? VT> took ~2h CPU time long, the wall time was ~7h when I run from the VT> head node, and ~4h when I run from another node, which is not the VT> head node. The reason why I avoided running from the head node was VT> that other users might use the head node to do their work, causing VT> the long wall time. Even when I did the test runs (short runs), VT> when no one used the cluster, the difference was still up to 25%, VT> depending on how many nodes I used. I do not know what I should do VT> to improve this difference. According to you, what should I look at VT> to fix the problem. (I also need to read the forum discussion as VT> Lorenzo said in his email). how should the number of users on the machine affect the performance? don't you use a batch system to regulate access to the nodes? cheers, axel. VT> VT> Thanks, VT> VT> Trinh VT> -- ======================================================================= Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu Center for Molecular Modeling -- University of Pennsylvania Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323 tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425 ======================================================================= If you make something idiot-proof, the universe creates a better idiot.
