May 28, 2025
—
in Ganymede, Ganymede 2, Juno, Storage, User Updates
Dear HPC users,
The HPC Scratch storage is a very high speed storage system intended for use during a job’s execution and can sustain high rates of IO. The purpose of this policy is to ensure that running jobs always have sufficient Scratch space during execution.
Policy
HPC Scratch storage (~/scratch directory) is for use during a job’s execution. User should utilize the Scratch space as “copy in, read and write within, copy out, clean up when done” space.
Data stored in Scratch is never backed up. Scratch will be purged regularly to maintain sufficient space for new batch jobs to run. When purged, any file or directory not accessed for 45 days will be deleted. The HPC staff will monitor Scratch utilization and, in the future, may reduce the (currently) 45 day time limit.
Above policy is available on the policies and guidelines page on the HPC website.
What you should do
We previously sent emails (attached) requesting all users to clean their Scratch directory. Please follow the instructions in that email and copy your needed data to other directories like home (~), work (~/work on Juno only) or group directories (/groups/<pi-name> on Juno, /mfs/io/groups/<pi-name> on Ganymede 2). If you need assistance with data migration, please submit an Atlas service catalog request under HPC Storage > Data Migration.
How to use Scratch
Please see the Scratch space section of the Juno directories and data management document for complete description of how to use Scratch space in your batch jobs. These instructions apply equally to Juno, Ganymede and Ganymede 2 users.
If you have any questions, please contact us through the Atlas service catalog under HPC Storage > General Storage User Support.
Thank you.
Rajesh Rao
Director HPC Operations
Office of Information Technology
Khalid Sarwar Warraich
Director HPC Facilitation
Office of Research and Innovation