Link Search Menu Expand Document
BIRC Computing Guide

High Performance (Cluster) Computing

Table of contents

  1. Priority Scheduling
    1. Getting access
    2. Submitting priority jobs
  2. Using GNU Parallel with SLURM

Priority Scheduling

BIRC has purchased a computing node (with 36 Intel Skylake cores) in the Storrs HPC. This node is available to all users of the cluster, but BIRC affiliates can be granted scheduling priority (relative to non-BIRC users).

Getting access

  1. Request a Storrs HPC account if you do not already have one.
  2. Have your PI email Roeland Hancock a list of NetIDs to be put on the priority user list. You will generally receive access within a few business days.

Submitting priority jobs

Include the following in your SLURM job description:

#SBATCH --partition=SkylakePriority
#SBATCH --account=roh17004

Using GNU Parallel with SLURM

GNU Parallel is a utility for easily distributing arguments across parallel jobs. A common use case is to process multiple subjects in parallel. For example, if you have have a list of identifiers in subjects.txt and script myscript.sh that takes a single identifier as an argument, you can process multiple subjects in parallel with the parallel command:

cat subjects.txt | parallel myscript.sh

parallel takes care of the details of running myscript.sh across processors or nodes. In this way, you can setup parallel jobs within a single SLURM job.

Getting parallel to work within SLURM adds some addition steps. The HPC team provides some utility scripts to make this easier. If you have trouble using these scripts, please contact HCP support.

Table of contents