running smufin on older openMPI

[hongen xu]

Hello, 

I am trying to run SMUFIN on clusters from a HPC center. They have old openMPI version installed (OpenMPI/1.4.3, order than the recommended 1.5.X). I prefer not to install openMPI locally, and it may interrupt the whole cluster. 

I am working on the chicken genome (genome size: 1GB) and sequencing depth around 18X. According the formula, it will use  (1000 * 18 * 2.3)*2= 82800MB memory. Since the job was arranged to different cluster/nodes randomly, the exact memory for each node can not be determined. I requested 40 nodes and 160GB memory (see my PBS file below), and I got my job killed by memory resource utilization exceeded requested capacity. 

My questions are 

1. is it OK to use older version of openMPI ? Should I adjust any parameters or not ?

2. should I request more nodes to solve the problem? 

thanks in advance,

kind regards,

Hongen

###########################################PBS file #######################

#!/bin/bash -login

 

### define resources needed:

### walltime - how long you expect the job to run

#PBS -l walltime=7:00:00

 

### nodes:ppn - how many nodes & cores per node (ppn) that you require

#PBS -l nodes=40:ppn=1

 

### mem: amount of memory that the job will need

#PBS -l mem=160gb

 

### load necessary modules, e.g.

module load powertools

 

### change to the working directory where your code is located

cd /mnt/home/hongenxu/smufin

 

### call your executable

mpirun -np 40 ./SMuFin --ref ref_genome/genome.fa --normal_fastq_1 normal_fastqs_1.S3 --normal_fastq_2 normal_fastqs_2.S3 --tumor_fastq_1 tumor_fastqs_1.S3 --tumor_fastq_2 tumor_fastqs_2.S3 --patient_id S3  --cpus_per_node 1 

qstat -f ${PBS_JOBID}

[SMuFin]

Hello Hongen,

The openMPI version must be 1.5.X (or higher).

About memory problem it should be solved, as you said, increasing the numbers of nodes.

Best regards,