SeaWulf gives you several places to keep files, each with a different size limit, backup policy, and purpose. Knowing which is which keeps your work safe and your jobs fast.
On this page: Storage areas · What goes where · Keep your data safe · Request project space · Check your usage
Storage areas
| Path | Size limit | File limit | Backed up? | How long files stay | Use it for |
|---|---|---|---|---|---|
/gpfs/home/<netid> | 20 GB | about 200,000 | Yes | Kept | Scripts, settings, small files, software you install |
/gpfs/scratch/<netid> | 20 TB | 10 million | No | Files older than 45 days are deleted | Job input and output while you're working on it |
/gpfs/projects/<group> | Up to 10 TB | about 10 million | No | Until the project ends | Data your group shares |
/gpfs/software/ | Managed by RCI | Kept | System-wide software (read only) |
All of these are the same on every login and compute node, so a file you save in one place is there for all your jobs.
What goes where
- Home is small but backed up. Keep code, job scripts, configuration, and anything you'd hate to lose here.
- Scratch is huge and fast but temporary. Run jobs here, then copy the results you want to keep somewhere safe. Never install software in scratch, since the purge will break it.
- Project space is shared with your group and lasts for the life of the project, but it isn't backed up.
Keep your data safe
To move data on and off the cluster, see File transfer from the CLI or, for large datasets, Globus.
Request project space
PIs can ask for shared project space when they request a project, or later through a ticket at iacs.supportsystem.com. Include:
- how much space you need,
- what the data is and how it will be used,
- how long you'll need it, and
- confirmation that your group understands project space isn't backed up.
Requests up to 10 TB are routine. Larger requests need a detailed justification and may be granted for a limited time. For long-term or archival storage, see the Research Storage picker.
Check your usage
myquota # space and file counts for home and scratch du -sh * | sort -h # what's using space in the current folder
More in Checking storage quotas. If you hit a limit, jobs can fail with "Disk quota exceeded", so check before big runs.