Titan/LSF quick start
Quick start on using LSF on the Titan Cluster at the Penn State University Department of Computer Science
Titan Quick Start
Accessing the cluster
Members of the I3C consortium may get access to the cluster by sending a request to cse helpdesk via email.
Once you have been added to the "titan" group you may ssh to the cluster.
**** Reminder All jobs submitted MUST have the "-P <research group>" option in the bsub. See below****
ssh <username>@titan.cse.psu.edu
What's available on the titan cluster?
- simics /home/titan/apps/simics/simics-3.0.31/
- ATLAS /home/titan/apps/ATLAS/
- openmpi /opt/openmpi/1.1.4/
- mpich /opt/mpich/gnu/
- intel mpi-rt /opt/intel/mpi-rt/3.0/
- infiniband libs /opt/ofed
- LSF job scheduler
What is different about the cluster?/Storage on the cluster.
titan cluster works differently as far as stored data goes.
The titan does not have direct access to the normal cse storage and has it's one temporary storage.
So what this means for you is you must copy your data to the cluster using "scp" and then move it off the cluster the same way. The space on the cluster is not for long term storage. It is not backed up. Do not keep important data there and do not use it for non research work.
Submitting a job to the cluster
Create a working directory under your home directory
mkdir ~/work cd ~work
Write and compile your code.
vi hw_hostname.c
#include <unistd.h>
#include <stdio.h>
int main()
{
char name[100];
gethostname(name, sizeof(name));
printf("%s says Hello!\n", name);
return 0;
}
gcc hw_hostname.c -o hw_hostname
Next create a LSF script that will process your job.
vi ~/work/submit.sh
#!/bin/csh
#BSUB -J serialjob # Name of the job.
#BSUB -o %J.out # Appends std output to file %J.out. (%J is the Job ID)
#BSUB -e %J.err # Appends std error to file %J.err.
#BSUB -P <your research group> # Set your Research group ID.
# ie: MDL, SPARSE, CPAC, MOBY, CITESEER, ALIPR, PARAGRAB, MCN, EMC2, ME, MATSCI
# ****Jobs found running without the Research Group ID will be killed.***
#BSUB -W 1:30 # Wall clock time of 1 hour and 30 min.
#BSUB -n 1 # Number of CPUs.
#
# Run serial executable on 1 cpu of one SMP node
~/work/hw_hostname
#End of Sample LSF Script
When your read you can submit your job using the "bsub" command
bsub < submit.sh
Jobs can be manipulated with the following command
- qstat to view jobs in queues
- qdel to delete job in queues

