[CIG-SHORT] Run on cluster using SLURM

[Matthew Knepley]

On Fri, Sep 28, 2018 at 3:57 PM Bing He <bing-he at uri.edu> wrote:

> Dear all,
>
> Is there anyone has successfully run Pylith using Slurm? Could you please
> give me some suggestions?
> I wrote the configuration file and run the code like this (attached). It
> can use many nodes and cpus, but the computing speed does not improve.
>

How big is the problem? Only bigger problems show any speedup

  https://www.mcs.anl.gov/petsc/documentation/faq.html#slowerparallel

Note also that absolutely no speedup in timestepping comes from parallelism.

Parallelism is really effective at running bigger problems. Strong speedup
(running
the same problem faster) is much harder and runs out very quickly.

  Thanks,

     Matt


> [pylithapp]
> scheduler = slurm
> launcher = mpich
> [pylithapp.slurm]
> shell = /bin/bash
> [pylithapp.launcher]
> command = mpirun -np ${nodes}
>
> pylith step06.cfg \
>     --launcher=mpich \
>     --scheduler=slurm \
>     --job.queue=general \
>     --job.walltime=2*day \
>     --job.name=speed \
>     —-nodes=40 —-scheduler.ppn=20 \
>     --job.stdout=log \
>     --job.stderr=err
>
> Thanks for your time and have a nice weekend!
> Bing
> _______________________________________________
> CIG-SHORT mailing list
> CIG-SHORT at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-short
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/cig-short/attachments/20180928/b638c861/attachment-0001.html>