[gmx-users] Haphazard Results with LAM-MPI
spoel at xray.bmc.uu.se
Wed Apr 9 08:17:01 CEST 2003
On Wed, 2003-04-09 at 06:13, Venkat Ramanan K wrote:
> We are a bunch of undergraduate students and have put up a cluster
> of 16 IBM's using LAM for the purpose of running GROMACS. During
> our simulation test runs we noticed that the running time
> increases as the number of nodes in the LAM topology increases. We
> couldn't diagonise the problem...plz help.
> We used the folowing commands:
> grompp -v -f full -o full -c after_pr -p speptide -np 16 -shuffle
> mpirun N -lamd -nger -nsigs mdrun -v -s full -o full -c after_full
> -g flog
Apart from the number of atoms, using the options nger and lamd will
make communicationparticularly slow. These option are meant for
debugging LAM application.
You want to run
mpirun -O -c2c -c N mdrun etc.
where N is the number of nodes.
Furthermore the network that you are using can have an influence. Is it
100 Mbit/s, is it switched?
Check the results on the benchmark web page. We have results for a
switched network and as long as you have a reasonably large system (and
no PME) it scales half decent at least.
> We got the following running times with GROMACS 3.1.4...
> 1 node - 13hrs 5:27
> 2 nodes - 16hrs 12:59
> 4 nodes - 17hrs 32:14
> 8 nodes - 45hrs 00:36
> 16 nodes - 45hrs 46:55
> Thanks in advance
> Odomos - the only mosquito protection outside 4 walls -
> Click here to know more!
> gmx-users mailing list
> gmx-users at gromacs.org
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
Dr. David van der Spoel, Dept. of Cell and Molecular Biology
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
More information about the gromacs.org_gmx-users