[gmx-users] mpi performance issues
Osmany Guirola Cruz
osmany.guirola at cigb.edu.cu
Mon Apr 19 23:37:01 CEST 2004
I have problems in my cluster with MPI "poooor scalable"
I think that my problem is in the options --with-tcp-short
And --with-rpi of my lam/mpi which should be the perfect values for
these options or how ca I found the prefect values for my network
I have a linux cluster Suse 8.2 dual PIII 933 100Mbs
From: David [mailto:spoel at xray.bmc.uu.se]
Sent: Monday, April 19, 2004 7:01 PM
To: gmx-users at gromacs.org
Subject: Re: [gmx-users] mpi performance issues
On Mon, 2004-04-19 at 22:47, istvan at kolossvary.hu wrote:
> I see there are a lot of threads on the mailing list on Gromacs mpi
> I also tried to do this "out of the box" and got similar, horrible
> as others. I read through the archive and got some clues, although it
> that scaling is extremely (molecular)system sensitive and I shouldn't
> much with PME. However, looking at the 20-50% CPU usage on dual
> nodes (3.06 GHz Xeons w/ 100Mbit Ethernet and Redhat 9.0) I was
> the payload, which is the size of the packets sent between nodes is
> small. I believe Erik has mentioned this once on the list. Apparently
> payload size is 1000 bytes (plus 40 bytes TCP/IP overhead) on our
> would be much more efficient having larger payload, say 1400 bytes. My
> question is whether this is something that can be set in Gromacs, or
> an OS parameter. In either case, how can I change this?
It is a feature of the MPI library. With LAM-MPI there is an option to
change this. Another option is to try the MPI-VIA implementation. Has
anyone tried that for GROMACS?
This should avoid the Linux TCP/IP stack, and have much lower latency on
cheap hardware. Volunteers please....
David van der Spoel, PhD, Assist. Prof., Molecular Biophysics group,
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596, 75124 Uppsala, Sweden
phone: 46 18 471 4205 fax: 46 18 511 755
spoel at xray.bmc.uu.se spoel at gromacs.org http://xray.bmc.uu.se/~spoel
gmx-users mailing list
gmx-users at gromacs.org http://www.gromacs.org/mailman/listinfo/gmx-users
Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-request at gromacs.org.
More information about the gromacs.org_gmx-users