Re: [modeller_usage] Modeller on SGE

[Modeller Caretaker <modeller-care@ucsf.edu>]

On 10/18/11 6:02 AM, Matías wrote:
> HI, I´m trying to run modeller in multiple cores on a Cluster with
> SGE. Do you have an example script to do this?

No, since it is highly dependent on your SGE setup. And it depends what 
you mean by "run modeller". There are two classes in Modeller for 
running on SGE systems: sge_qsub_job and sge_pe_job. The first uses 
'qsub' from within the Modeller run to start an SGE job to do the 
computation (typically you run Modeller on your head node or some other 
machine which can 'qsub' to your cluster). The second is designed to be 
used in a Modeller job which is running on your cluster in a PE; it will 
allow Modeller to use all the nodes in that PE.

A basic usage of the first for comparative modeling would be something 
like (see also examples/automodel/model-parallel.py):

from modeller import *
from modeller.automodel import *
from modeller.parallel import *

j = sge_qsub_job(maxslave=10)

env = environ()
a = automodel(env, alnfile=...)
a.starting_model = 1
a.ending_model = 100
a.use_parallel_job(j)
a.make()

This will automatically 'qsub' an SGE array job containing 10 tasks when 
you call a.make(), so that the modeling is split between 10 processors.

See also
http://salilab.org/modeller/9.10/manual/node438.html

	Ben Webb, Modeller Caretaker
--
modeller-care@ucsf.edu             http://www.salilab.org/modeller/
Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage