[gmx-users] jobs in parallel

Chng Choon Peng cpchng at bii.a-star.edu.sg
Wed Aug 11 02:44:03 CEST 2004


Hi Patricia,

  I might be able to offer some help, since I also run GROMACS on a cluster
with Myrinet (though not of comparable size as yours :)

You have to supply a machinefile to mpirun which otherwise starts your jobs
on the default node list (aka node01, node02, say).

Example:

#PBS -l nodes=2:ppn=1

NPROCS=`wc -l < $PBS_NODEFILE`

/usr/local/mpich-gm/bin/mpirun -machinefile $PBS_NODEFILE -np $NPROCS
/usr/local/gromacs3.2.1/i686-pc-linux-gnu/bin/mdrun_mpi -s topol.tpr

Hope this helps.

cheers,
Choon-Peng

-- 
Mr. Choon-Peng CHNG
Research Associate
Computational Biology Group
BioInformatics Institute, BMSI, A*STAR
30 Biopolis Street
#07-01 Matrix Building
Singapore 138671
Tel (O): +65 64788301 Fax (O): +65 64789047
www.bii.a-star.edu.sg/~cpchng


On 8/11/04 7:54 AM, "Patricia Soto Becerra" <psoto at ultra.chem.ucsb.edu>
wrote:

> 
> Hi gmx-users!,
> I have some difficulties to run parallel jobs in a new cluster (128 dual
> CPU Xeon 3.06 Ghz processors with a Myrinet high-speed interconnect).
> 
> When I submit my job:
> 
> mpirun -np 2 mdrun_mpi -s topol.tpr
> 
> It is reported to be running on a specific node. For example, when I
> type *qstat* it is reported to run on node10.
> 
> If I check the md0.log and md1.log files, the processes are running on
> other nodes (say node01 and node02)!!. When I connect to the nodes
> reported in md0.log and md1.log indeed the jobs are running there (node01
> and node02), and nothing in node10 (the one that is reported by
> *qstat*). The performace is so low under these conditions that for any
> benchmark of gromacs the parallel jobs turn to be a lot slower than a
> single node job.
> 
> My guess is that there is a problem with PBS (?) but at the moment I
> cannot find a solution.
> 
> Does anybody have any idea on how to solve this???!
> 
> Patricia.
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Dr. Patricia Soto
> Post-doctoral Researcher
> Department of Chemistry and Biochemistry
> University of California, Santa Barbara CA 93106 - 9510
> United States
> 
> **Como bien nos ensena la vida, la verdad es una aspiracion inalcanzable
> que, en todo caso, habita al otro lado del muro que forman las
> innumerables verdades que cada uno sostenemos por convencimiento o por
> intereses mas o menos legitimos**  Jesus A. Nunez Villaverde, 03-08-04
> 
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> _______________________________________________
> 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 mailing list