Skip to content

Latest commit

 

History

History
56 lines (37 loc) · 3.16 KB

README.md

File metadata and controls

56 lines (37 loc) · 3.16 KB

Getting Started with the Research Computing Clusters - Running Example Jobs on the HPC Clusters

About

This resource provides example jobs that can be run on Princeton University's Research Comptuing clusters.

Useful links

Running Example Jobs on the HPC Clusters

The sample jobs above are written with Nobel, Adroit, Della, Stellar, and Tiger in mind. You will probably need to use different environment modules on Traverse.

Follow the directions below to begin running simple jobs on Adroit. After SSHing to Adroit the first step is to cd (change directory) to your directory on /scratch/network/<YourNetID>. We do this because /scratch/network is a much faster filesystem with more space than /home.

ssh <YourNetID>@adroit.princeton.edu  # vpn required from off-campus
cd /scratch/network/<YourNetID>  # on tiger, della or stellar replace /scratch/network/ with /scratch/gpfs/
git clone https://github.com/PrincetonUniversity/hpc_beginning_workshop
cd hpc_beginning_workshop

Then choose an example and follow the directions. For instance:

cd python/cpu
cat README.md

For more detailed instructions on running an example Python or R job, see the First Slurm Job section of the Guide to the Princeton Clusters.

Where to store your files

To familiarize yourself with the cluster's file systems and where to store to your files, review our Data Storage page.

IMPORTANT: You should run your jobs out of /scratch/network on Adroit and /scratch/gpfs on the other clusters. These filesystems are very fast and provide vast amounts of storage. Do not run jobs out of /tigress or /projects. These filesystems are slow and should only be used for backing-up the files that you produce on /scratch/gpfs or /scratch/network. Your /home directory on all clusters is small should only be used for storing source code, executables, Conda environments and small data sets.