Storage Options#
DesignSafe gives you three main storage systems, each designed for a different phase of the research lifecycle.
Corral is the long-term home for your data. It lives on a networked storage system at TACC, with automatic backups and strong support for collaboration and publication. Corral is accessible from almost everywhere in DesignSafe — the Data Depot, JupyterHub, VMs, and via Tapis Apps. Think of Corral as your research “filing cabinet”: it’s where data is safely stored, shared with collaborators, and archived for future reuse.
Work is mounted directly on compute systems for active projects, so it offers high performance but without backups. It’s ideal for staging inputs before you run jobs and holding outputs afterward.
Node-local storage is attached to the compute node your job runs on. It’s ephemeral and disappears when the job ends, but it’s extremely fast and useful for scratch space during computation.
Here’s a compact table that shows at a glance how Corral, Work, and Node-local differ in persistence, performance, and access.
Storage Type |
Persistence |
Performance |
Access From |
Best Use Case |
|---|---|---|---|---|
Corral |
Long-term, backed up |
Moderate (network) |
Data Depot, JupyterHub, VMs, Tapis |
Archiving, collaboration, publication |
Work |
Long-term (not backed up) |
High (on system) |
Compute systems, Data Depot, JupyterHub |
Staging input/output files for jobs |
Node-local |
Temporary (deleted at job end) |
Very High (local) |
Only during active compute job |
Fast scratch I/O during runtime |
You can immediately see why Corral is for long-term use, while Work and node-local are for jobs.
This table makes the tradeoffs clear:
Corral = safe, shared, persistent → but slower.
Work = faster, system-mounted, but not backed up.
Node-local = fastest, but ephemeral.
Prepare in Corral → Run in Work/Node-local → Archive back to Corral!!