Start here if you are new to Datahub and the Platform for Applied Research, Analytics, Science and Open Learning (PARASOL, formerly DSMLP). Find out what the service offers every course and lab, the constraints that apply to all work on it, and the page to begin with for your role.
Datahub and PARASOL¶
UC San Diego's Datahub and the Platform for Applied Research, Analytics, Science and Open Learning (PARASOL) provide students, instructors and researchers with Jupyter, RStudio and other computational resources, for scheduled classes, formal independent study, student projects and research. You reach PARASOL through Datahub in a browser, or through the login node from a terminal. PARASOL was formerly DSMLP, and research work on it was formerly called the Research Cluster. Older pages and articles use both names.
The platforms are used in courses ranging from introductory Data Science lectures to graduate-level applied machine learning, and in curricula in Biology, Music, Social Sciences, and Public Health. In AY 2025-26, the platforms hosted more than 140 classes, 75 instructors, and 18,000 student enrollments.
Datahub is the web interface, at datahub.ucsd.edu. PARASOL is the cluster that runs Datahub sessions, and it is also reachable from a terminal. Datahub and PARASOL are a single service, and the same files are present in both.
Where to Begin¶
Each audience has its own starting page.
| Audience | Starting page | Assumes |
|---|---|---|
| A student enrolled in a course that uses Datahub | Using Datahub in a Course | A web browser and nothing else |
A student in such a course, where the course requires SSH, launch.sh, batch jobs, VS Code, or a GPU class the course menu does not offer |
Working from the Command Line | A terminal and ssh |
| Teaching or assisting with a course | Teaching with Datahub and PARASOL | A browser for most tasks |
| Undertaking a personal project, independent study, capstone, or club or team project | Projects & Independent Study | Familiarity with a shell |
| A graduate student, postdoc, undergraduate, or staff researcher working without a lab workspace | Research on PARASOL | Familiarity with a shell |
| Faculty arranging access for a research group | Setting Up a Research Lab | Some steps are performed by ITS staff |
Coursework and a personal project are two separate workspaces, with different eligibility, different resources, and different access lifetimes. Using Datahub in a Course and Working from the Command Line cover coursework. Projects & Independent Study covers a personal project.
Workspaces¶
Users are divided into groups called workspaces: one per course, one per lab, or one general catch-all population. A workspace anchors all cluster configuration: rosters, storage, container images, GPU access, and storage and GPU quotas. See What a Workspace Is and What It Controls.
A member may belong to several workspaces at once, for example as a TA for one course, a student in another, and a member of a lab, and chooses the workspace to work in at launch. A member of more than one workspace has a home directory in each. See Belonging to Several Workspaces.
See also: Workspaces & Storage
Standard Features¶
Job Environment¶
All jobs, whether browser-based Jupyter sessions, command-line jobs, or batch jobs, run in a compute environment derived from the member's account and workspace together with any job-specific configuration. A course session in the browser typically begins at 2 CPU cores and 4GB RAM. See The Browser Session. Upper limits are based on class size, demand, and capacity.
GPUs¶
GPUs are organized into five classes by size rather than by hardware model. Each workspace is given access to one or more classes, matching its anticipated work. GPU Class Sizes lists the classes. Workloads by GPU Class gives examples of the work each class suits.
Storage¶
Each member has a per-user, per-workspace home directory; a shared area readable by everyone in the workspace; and a per-user, cluster-wide private area available in every workspace. Where Files Live lists the directories. Workspace and Personal Quotas gives the quotas for course and research home directories. Optional mounts of external storage are available. See Mounting External Storage.
Standard Software Images¶
Curated images cover the widely used Python, R, and Julia data analysis stacks. They include a CUDA/GPU-enabled image with TensorFlow and PyTorch, and an RStudio image. Courses and labs may further customize or pin their own images. Standard Images, Tags, and Pinning describes the images.
Ways to Work¶
Work is done in a browser, at a terminal over SSH, or in a desktop editor such as VS Code connected to a cluster container. Access describes each route.
GPU Reservations¶
From Fall 2026, access to GPUs is managed by a reservation system. A booking, made in the reservation app at reserve.dsmlp.ucsd.edu, holds capacity for a specific window, and for the length of that window sessions are admitted ahead of the walk-up queue. Ad-hoc, on-demand use without a booking is also supported. Reservations describes booking. Reservations, per-group quotas, and Service Unit budgets are the mechanisms that manage contention for GPUs. See GPU Access.
Warning
An on-demand launch draws on the Service Unit budget as a booked window does, and launching an eligible session authorizes that spend. See On-Demand Lease Charges.
Platform Constraints¶
Root Access¶
Containers run unprivileged, under the member's own UID, with no root or sudo
inside the container. See
Root Access and System Packages.
sudo apt-get install fails by design, and the failure is not a fault to
report.
Computation on the Login Node¶
The login node is for launching jobs and moving files, and running computation on it is prohibited. See What the Login Node Is For.
Idle GPU Sessions¶
An idle GPU session still holds its GPU, and the session is reclaimed once it has been idle long enough. See What Counts as Idle.
Support¶
Questions and feedback go to the contact for the topic.
| Topic | Contact |
|---|---|
| A course a student is enrolled in | The instructor or TA, first |
| Datahub or PARASOL itself | parasol@ucsd.edu, or the ITS Service Desk |
| Research work, or Universal Scale Storage | rcd-support@ucsd.edu |
| Which platform is appropriate for a research project | Research IT |
| 1:1 Consultation, for instructors, TAs, and Technical Points of Contact (TPOCs) | ucsd-datahub.youcanbook.me |
Urgent or broadly scoped problems may be escalated through the ITS Service Desk. A ticket for such a problem states what is affected and how many people. See Incidents Affecting More Than One Person. Response Targets gives ITS response targets. Support & Technical Consultation sets out the terms of 1:1 Consultation. Getting Help lists all support routes.
Policy¶
Policy sets out the conditions of use.
- Highly sensitive data (P4), such as clinical records or export-controlled information, must not be used on Datahub or PARASOL, and legally or contractually protected information (P3) may be permitted after review. See Data Classification.
- Datahub and PARASOL system resources are shared among all courses and research groups, and demand, in particular for GPUs, may exceed capacity at peak hours during 10th and Finals Weeks or at assignment deadlines. See Use of Shared Resources.
- Datahub may be unavailable during scheduled maintenance for time-sensitive updates or security patches. Quarterly maintenance ends every running job. Scheduled Maintenance gives the schedule.
- Datahub and PARASOL are not to be used to host externally available services or applications except as required for coursework or projects. See Hosting Externally Available Services and Availability and Reliability.
- The campus-wide IT policy on acceptable use applies to Datahub and PARASOL. See Appropriate Use.