The College of Letters & Science IT (CIT)
SSMS | Life Sciences | Santa Barbara, CA 93106-4020
cit.ucsb.edu
- Service Name: Ephemeral Computing via BinderHub
- Service Category: Instructional and Research Support
- Service Type: Instruction & Self-Service Computing
- Service Owner: Andreas Boschke
- Business Owner: Ted Cabeen
- Last Revised: August 3, 2026
To provide users with a fully dynamic, self-service computing environment where custom environments are built and launched on demand directly from git repositories. This enables instructors and researchers to create, manage, and share reproducible execution environments without manual administrative provisioning.
All UCSB community members who manage their own computing environments via git repositories (e.g., Forgejo, GitHub, GitLab) and require temporary, reproducible execution environments for coursework, workshops, or data analysis related to academic research.
- Git-Driven Environment Build: Automatic creation of executable JupyterHub environments directly from git repository configurations (e.g.,
environment.yml, requirements.txt, Dockerfile, APT.txt).
- Self-Managed Images: Complete control by users over software packages, versions, and dependencies via their own git repositories without needing CIT ticket approval.
- Ephemeral Sessions: Instantly launched, temporary compute instances designed for interactive analysis, code execution, and demonstration.
- Instant Sharing: Direct URL-based sharing allowing any user with the link to launch the identical environment from the source repository.
- Multi-Repository Support: Launch an unlimited number of distinct environments based on different repositories or git tags/commits.
- Authentication: Integrated login using standard UCSB NetID credentials for launching instances.
- No Persistent Data Storage: All instance state, local changes, and created files are completely ephemeral and will be destroyed upon session termination or inactivity timeout.
- User Data Management: Users are solely responsible for pushing any code, results, or data back to remote storage (e.g., git remote repositories, cloud storage, local device via browser) before ending their session.
- Resource Limits: Compute instances operate under strict CPU, RAM, and session timeout limits to ensure fair resource allocation across the shared cluster.
Rather than requiring administrative provisioning for every environment, instructors and researchers can build their custom BinderHub instances using pre-configured, maintained base images as starting templates. All of UCSB's single-user Jupyter images are publicly available on GitHub via the UCSB-PSTAT Organization. These can be used directly or referenced as parent images in your repository's configuration (Dockerfile, environment.yml, etc.). Example of some popular images are:
- Base Environment: jupyter-base — Minimal, lightweight starting image for custom Python setups.
- SciPy: scipy-base — Pre-loaded with standard Python data science stacks (pandas, scikit-learn, matplotlib).
- R & RStudio: base-rstudio — Pre-configured environment supporting R Notebooks and native RStudio interfaces.
- Distributed Computing: all-spark-base — Includes Apache Spark integration for large-scale data processing (PySpark, SparkR, Scala).
| Metric |
Details |
| Hours of Availability |
24/7/365 general access for students, instructors, and researchers. |
| Supported Hours |
UCSB Business Hours. Off-hours support is provided on a best-effort basis. |
| Target Uptime |
99.9% outside of maintenance windows, within supported normal business hours. |
| Maintenance Windows |
Standard system maintenance (OS updates, cluster deployment upgrades, maintenance rebuilds) is scheduled in the windows between terms—specifically from the last day of instruction of a term to the first day of instruction of the following term. The service may be intermittently unavailable during these inter-term periods. |
| First Response Time |
Within 1 business day. |
| Issue Resolution Time |
Within 2 business days for core service issues. |
- Data Retention: NONE (0 Days). Local storage is strictly temporary. Files modified or created inside a BinderHub session are automatically deleted when the container shuts down.
- Recovery Point Objective (RPO): Not Applicable for local user data (data must be stored externally by the user). Build image cache RPO is 24 hours.
- Recovery Time Objective (RTO):
- Service Restoration: 48 hours (reinstatement of the core BinderHub cluster and image registry).
- Data Restoration: Not Applicable (no persistent user volume recovery is supported).
- Data Protection Level (IS-3): P1 / P2 (Unsuitable for sensitive or protected data P3/P4 due to non-persistent and public link share design)
- Data Availability Level (IS-3): A1
- Data Recovery Level (IS-12): RL0 (No data recovery provided)
- Cost: Covered by the Student Tech Fee and CIT IUC Fund (free to instructors and students).
- How to Access: Log in via the binderhub.
- Additional Support: Submit a ticket via CIT Support.
- JupyterHub Service: CIT-managed persistent environments for structured courses requiring long-term student data retention and pre-configured image distributions.
- Container Registry & Caching: Internal image registry to cache previously built repository images for rapid subsequent launches.
- Automated Image Build Engine: Background build service that parses repository configuration files to generate OCI-compliant container images on demand.