Data & Storage

ACCESS offers several kinds of storage and several ways to move data. The right choice depends on the size, lifetime, and use of your data.

Storage Types and Policies

Every compute allocation includes local file-system storage. Additional archival and stand-alone storage can be requested separately.

Resource file-system storage

Local disk and scratch space included with every compute allocation. Includes $HOME and $SCRATCH.

Archival storage

Large-scale persistent storage, usually requested alongside a compute allocation. Usually not backed up.

Stand-alone storage

Storage independent of a compute allocation. Useful for data hosting and preservation projects.

Allocatable Archival and Stand-alone Systems

	Expanse (SDSC)	Ocean (PSC)	Ranch (TACC)
System type	Medium-term disk	Persistent disk & tape hybrid	Long-term tape archival
Stand-alone?	No	No	Yes
File system	Lustre	Lustre	Lustre
Default / max space	500 GB / 75 TB	500 GB / none	none / 1 PB
Retention after expiry	3 months	3 months	6 months

Each resource name links to its documentation page.

File-system Types

Storage tier	Typical size	Backed up?	Auto-purge?	Best use
`$HOME` — permanent, small	≤ 50 GB		—	Scripts, configs, small result files
`$SCRATCH` — large, temporary	1–20 TB	—		Active job input/output data only
Archival — long-term, large	Up to 1 PB	Varies	—	Long-term retention, infrequent access

$SCRATCH is not backed up and is purged regularly.

Move results you need to $HOME or off-system storage after your job completes. Check the purge schedule for each resource — see the Resources listing.

How do I check my quota?

Check your quota before submitting jobs to avoid failures mid-run. Most systems support:

$ quota

Some resources provide a custom command such as myquota. Check the resource's user guide. See the Resources listing for per-resource details.

What are the backup schedules?

Backup policies vary by resource and file-system tier. $HOME directories are typically backed up daily or weekly. $SCRATCH and $PROJECT spaces are usually not backed up — they are intended for active work only. Archival storage (where available) may or may not include snapshots. Check the resource-specific documentation or contact the resource provider to confirm the exact schedule for the system you are using.

What happens when $SCRATCH files are purged?

Files in $SCRATCH that have not been accessed for a set number of days (typically 30–90 days, depending on the resource) are automatically deleted without warning. Purged files cannot be recovered. Move any results you want to keep to $HOME, $PROJECT, or your local storage as soon as a job completes. You can check the purge policy for a specific resource in its documentation or by contacting the resource provider.

Can I share files with collaborators?

Yes. The recommended method is Globus Sharing — it lets you grant read or read-write access to a path on any Globus-enabled system without requiring your collaborators to have an account on that resource. See Share Data on ACCESS Systems for setup instructions. For simpler cases within the same resource, standard POSIX group permissions (chmod, chgrp) also work if your collaborator has an account on the same system.

How do I request a storage increase?

Storage increases are handled through the ACCESS allocations process. Log in to the allocations portal (allocations.access-ci.org) and submit a storage supplement request for your existing allocation. Include your current usage, why you need more space, and how long you need it. For urgent situations, open a help ticket and reference your allocation number — the resource provider may be able to grant a temporary increase while a formal request is processed.

File Transfer

The right tool depends on the size of your data and whether you need to transfer between two remote systems. Supported methods, data-transfer-node hostnames, and Globus endpoint names vary by resource — see the Resources listing.

Globus Recommended for most users

Reliable, high-performance web and command-line interface for moving data across ACCESS resources and to and from your local system. Transfers resume automatically if interrupted. Free for ACCESS users.

Rclone

Command-line tool for cloud storage and remote systems. Good for scripted or automated transfers.

rclone copy source:path dest:path

Cyberduck / FileZilla / Bitvise

Graphical SFTP clients for users who prefer a point-and-click interface for SFTP connections.

scp / sftp

Built-in command-line tools. Available on any Linux or macOS system without installation.

scp local_file.dat username@resource.edu:~/data/

Data Analysis

All ACCESS HPC, high-throughput, storage, visualization, and cloud resources can be used for data analysis. GPUs are useful for AI/ML-accelerated analysis.

Find analysis software The Software Documentation Service lists every tool available across ACCESS resources, searchable by resource, discipline, and license type. Domain-specific analysis Science Gateways offer ready-to-use analysis environments tuned to particular research communities — no terminal required. Pre-staged datasets Many resources maintain genomics, geospatial, meteorological, and AI training datasets already on-system — no transfer needed.