Discussion Forum

Dear.

When I submit a computation task in the multi-nodes cluster with #PBS commands, I find that it took more time to compute with two-nodes than with one-nodes.

Here, one-node contain 16 cores, and when i compute with two-nodes, all 32-cores are used. One-node need 1141s to complete while two-nodes need 1330s.

Could you tell why? Is the problem with my submit script or comsol itself?

Thank you very much!

My #PBS script as follows

#PBS -j oe

#PBS -q "standard"

#PBS -d "/home/xxx/multiphysics/bin/"

#PBS -o "/home/xxx/logs/pbs.out"

#PBS -e "/home/xxx/logs/pbs.err"

#PBS -l nodes=2:ppn=16

#PBS -l walltime=1200:00:00

#PBS -q queue 2

#PBS -S /bin/bash

###############Load environment

module load intel2018

module load anaconda3

module list

##############Get computing resource variables

cd $PBS_O_WORKDIR

NP=cat $PBS_NODEFILE | wc -l

NN=cat $PBS_NODEFILE | sort | uniq | tee /tmp/nodes.$$ | wc -l

cat $PBS_NODEFILE > /tmp/nodefile.$$

###############Execution code

comsol batch -nn 2 -nnhost 1 -np 16 -inputfile /home/xxx/test.mph -outputfile /home/xxx/test-solved.mph -batchlog /home/xxx/test.mph.log

############## Clean up temporary files

rm -rf /tmp/nodefile.$$