[CSC 435] Give it shot...

[Andrew J. Pounds]

Okay guys --  it's not perfect, but it is functional and in its limited 
capacity works great.  I am speaking of the new batch resource manager 
and scheduler on zeus.  I submitted over 200 jobs this morning and all 
but two completed successfully.  To give it a shot create a pbs script 
file.   For example, I have a pbs script file called testing.pbs whose 
contents look like.

#PBS -N Matrix_Test
#PBS -l nodes=1:olympus:ppn=1,walltime=3:00:00
#PBS -m abe
#PBS -j oe
#PBS -k n
#PBS -V
cd /home/pounds/matrix_testing
mm >results.dat

When run on zeus, this will select 1 node from the cluster, allocating 3 
hours of walltime (the default it 1 hour if you do not specify this) and 
then execute the commands below the last #PBS directive.  In this case I 
am running an executable called mm and putting the results in a file 
caller results.dat

Once you submit the job you can check the job status with the command 
/usr/local/bin/qstat or /usr/local/maui/bin/showq

Both commands provide slightly different information.   If the job runs 
correctly, you should get  file name Matrix_Test.oXXX
where the XXX refer the job number.

There is still quite a bit of tweaking to do -- and I have only created 
one queue (called batch) -- but this should at least let you do timing 
runs of you matrix calculations while away from the machines.   There is 
still an issue with using the batch scheduler in "interatcive" mode (the 
-I option) -- so don't use that until I give the all clear sign.  Also, 
don't allocate more than one node yet.  Still working on the details of 
"sharing"...




-- 
Andrew J. Pounds, Ph.D.  (pounds at theochem.mercer.edu)
Associate Professor of Chemistry and Computer Science
Mercer University,  Macon, GA 31207   (478) 301-5627