[torqueusers] PBS in Cluster

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Mon Feb 1 07:29:31 MST 2010


On Sat, Jan 30, 2010 at 7:58 PM, Tim <timlee126 at yahoo.com> wrote:
> Hi,

tim,

> I am learning and have some questions about using PBS to submit jobs in a cluster.
>
> (1) If not using qsub to submit a job, will the job be running only on the single node where it is submitted?
>
> Even if the job is parallelized by MPI and run by mpirun, is it still running on the single node, not the others?
>
> So is qsub used for submit job to running on other nodes in the cluster?

qsub will ask to reserve nodes but it knows nothing about MPI.
similarly mpirun can be set up so that it will as the batch system
about which nodes are reserved for it, but this is not needlessly so.
you can use mpirun interactively and without batch/qsub by providing
a host list. some MPI libraries do need this always. hence $PBS_NODEFILE.

>
> (2) In a pbs script that is submitted by qsub, are all the commands executed one after the other?

it is a regular shell script.

> If I want to run several executables at the same time, is it to make these calls background by adding "&" at the end?

if you background them, they will execute in parallel. if you need
them in sequence,
you can try using job dependencies (-W afterok:####).

> (3) In the pbs script, if the several calls to run the executables are running the same time by running background, are the number of nodes and processors per nodes specified to be the total needed by all those calls?

the script submitted to qsub will run as a regular shell script on the
first(!) host assigned to the job.
no job distribution or limitation is done. exectly this is why one
would submit each execution in
separate jobs. if you use OpenMPI, you can consider using an appfile
to scatter multiple jobs
across the allocated nodes, but keep in mind, that when those jobs
take different amounts of
time, you'll have idle nodes that cannot be used by other jobs.

> If yes, and the number of nodes and processors per nodes specified are not completely available but enough to run some of the calls, will some of these calls be run first or delayed until the requested total resources are completely available?

again, this kind of load balancing is exactly what batch systems in
itself are for.
just submit all requests in individual qsub statements and all this will happen.

cheers,
    axel.

> Thanks and regards!
>
>
>
> _______________________________________________
> torqueusers mailing list
> torqueusers at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torqueusers
>



-- 
Dr. Axel Kohlmeyer    akohlmey at gmail.com
Institute for Computational Molecular Science
College of Science and Technology
Temple University, Philadelphia PA, USA.


More information about the torqueusers mailing list