Advanced Research Computing (ARC) - shannonhouck/mayhall-lab-manual GitHub Wiki
Introduction
The Advanced Research Computing (ARC) systems at Virginia Tech are a group of supercomputer clusters. We use them for most of our calculations.
PBS Job Scheduler
A job submission script contains information about the following:
- Walltime β The amount of time allocated for the job. This is given in the format
DD:HH:MM:SS
, whereD
is days,H
is hours,M
is minutes, andS
is seconds. - Nodes β The number of nodes (computers) given to the job.
- Processers Per Node (ppn) β The number of processors required on each node.
- Memory (mem) β The memory required for the job
- Queue (q) β Which queue to use. Information about the different available queues can be found on the ARC website. (We usually use the nmayhall_lab queue.)
- Allocation (A) β Which allocation to take funds from.
- Group (W) β Which group you belong to.
Sample PBS Submission Script
A sample submission script is shown below:
#PBS -l walltime=00:00:10:00
#PBS -l nodes=1:ppn=24
#PBS -l mem=120GB
#PBS -q nmayhall_lab
#PBS -A qcvt_doe
#PBS -W group_list=nmayhall_lab
# this is where we do things
echo βHello, world!β
exit;
This is saved as a shell script and can be submitted to the queue as follows:
$ qsub submit.sh
Output and error files will appear when the job ends. Some useful variables are:
$PBS_O_WORKDIR
: Work directory in which theqsub
command was executed.$TMPDIR
: Temporary directory created on the node. This is deleted at the end of the job, so any files you need from here must be copied back to somewhere permanent.
More information about storage on the ARC systems can be found here.
SLURM Job Scheduler
[ Under Construction!! ]