Re: PBS & local (scratch) space

From: Alessandro Cembran (cembran_at_chem.umn.edu)
Date: Wed Sep 20 2006 - 16:19:19 CDT

I think that the best way would you to ask to your system administrator, since it is more a system-dependent issue than anything related to NAMD in particular. Once you figure out how to do that, you can use a similar pbs script for any program, not just NAMD. Anyway, this is how I do it here, but as I said, things might be different in your cluster:

#PBS -l nodes=1,ncpus=.....,mem=.....,walltime=.....
#PBS -m abe

# Define the directories
export TARG_DIR=/scratch/$USER/$PBS_JOBID

# Create a directory for me in /scratch on the compute node if it does not exist
if [ ! -d /scratch/$USER]; then
    mkdir /scratch/$USER;
fi

cd YOUR_JOB_DIRECTORY

cp -pr * $TARG_DIR

cd $TARG_DIR

EXECUTE NAMD

cp -pr $TARG_DIR WHERE_YOU_WANT_YOUR_FINISHED_JOBS_TO_BE_PUT


Anyway, I don't seem to recall any significant speedup in doing that on the altix we have here, at least when you save your trajectory every few hundreds steps. It's anyway safer to do that, so that if your home crashes or there are problems in the nework, your job can still keep running.

Alessandro


Arturas Ziemys wrote:
Hi,

Few months ago I posted a question about scratch space and NAMD on cluster without answers. I just wanna check if somebody MAYBE now have ana advice or clue ...

How to ensure that NAMD job using PBS will be submitetd to the local (scratch) space of nodes instead of using user disk space ?

Best
Arturas


  

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:36 CST