Compute Node Local Work Directory

BC Project: FY20-01
Date of Policy: 28 Aug 2020
Last Updated: 05 Feb 2026 (see Revision Log)

This policy defines a common method by which a user's batch job can access a unique work directory that is local to certain types of compute nodes on some allocated HPC systems. Systems may implement the directory in different ways such as SSD, NVMe, Optane memory, and high-speed HDD devices. This local work directory provides enhanced performance characteristics to application reads and writes.

The environment variable $LOCALWORKDIR defines the path to the directory for the batch job on the nodes assigned to that job. The directory's contents will be deleted (by the system) after the batch job ends so that its full capacity can be guaranteed to the next job/user on the nodes. Since the workspace defined by $LOCALWORKDIR is only available for the duration of the batch job, the user is responsible for copying any files that need to be retained to another space such as $WORKDIR, $HOME, or $ARCHIVE_HOME. Likewise, at the beginning of a job, files that require high bandwidth can be copied to $LOCALWORKDIR from these other spaces.

Each center/system can define how best to implement the $LOCALWORKDIR deletion requirement.

Shared compute nodes are exempt from this policy.


Revision Log
Date Revision
05 Feb 2026BC Team Audit. Title Change.
16 Dec 2021BC Team Audit
28 Aug 2020Initial version