We provide scratch space on our parallel filesystem for writing large amounts of temporary during running jobs. Scratch space is provided on a first-come, first-serve basis, and there are no individual user quotas.
Data in the scratch volume is periodically deleted by an automated script. The scratch volume is not for long-term data storage.
What is scratch space for?#
Sometimes a compute job may generate large amounts of temporary data while the job is running, but that data is then discarded when the job completes. Our scratch volume exists to accommodate this type of data.
If you expect that your job will generate more data than is allocated in your home directory or research group storage quota, you may request temporary scratch space.
Scratch space etiquette#
Scratch space is intended to be temporary storage for reading and writing for running jobs. Once your jobs complete, you should move any data you have in the scratch volume to a permanent location (either on our storage system or your own device).
Scratch data will be periodically deleted. Since scratch space is a shared resource, we reserve the right to delete any data at any time that inhibits the ability of other RCC users to access the volume.
Currently, we provide scratch space on a per-request basis. Any RCC user can request access to scratch space by sending us a support request.
Using scratch space#
If you have been granted access to scratch space on our storage system, then you can locate it as follows:
For example, HPC user abc12a:
For convenience, you may wish to create a symlink (shortcut) inside your home directory to your scratch space:
You can now configure your HPC jobs to read/write to your scratch directory during execution.
Checking scratch volume storage quota#
You can check the storage usage and quota in the scratch volume anytime using the following command:
This output shows that there are 6.8TB in use out of 15TB available.