Draft

This documentation is in draft and under active review. Figures for the Fall 2026 GPU reservation system are not yet published, and some pages describe behavior that has not been verified against a primary source. Confirm any figure with datahub@ucsd.edu before relying on it.

Datahub & DSMLP Documentation

This is the documentation set for UC San Diego's Datahub and Data Science & Machine Learning Platform (DSMLP). It is organized by audience and by subject area.

Start Here

Each audience page is listed with its intended reader and what it assumes.

Page Intended for Assumes
Datahub and DSMLP Overview Everyone, with routing to the page for each audience Nothing
Using Datahub in a Course Students enrolled in a course that uses Datahub A web browser
Working from the Command Line Students in a course that requires a terminal A terminal and ssh
Teaching with Datahub and DSMLP Instructors and TAs A browser; a shell for course-specific customization
Projects & Independent Study Student projects, independent study, capstones, clubs and teams Familiarity with a shell
Research on DSMLP Researchers working without a lab workspace Familiarity with a shell
Setting Up a Research Lab Faculty provisioning access for a group Some steps are performed by ITS and Research IT

Technical Documentation

Technical detail that applies to more than one audience is organized by subject area. The audience pages link to it.

Subject Covers
Access Datahub in the browser, the login node, remote editor setup, sign-in and session problems, and when access starts and ends
Workspaces & Storage Workspaces and what they control, directories and quotas, moving and sharing data, and datasets
Running Jobs launch.sh options, job modes and the runtime limit, monitoring a running job, checkpointing and logging, and direct Kubernetes use
GPU Access GPU classes, reservations, Service Units and budgets, quotas and availability, and the causes of a session ending
Environments Standard images and pinning, customizing an environment, and building a custom image
Grading Grading tools and interfaces, the nbgrader workflow through Canvas import, and common grading failures
Reference Error messages, support routing, the glossary, HPC vocabulary, group management, policy, and software

Support

Questions and feedback about this documentation, or about Datahub and DSMLP as a whole, go to the following contacts.