Wake Forest University's Distributed Environment for Academic Computing (DEAC) Cluster is a centrally funded, enterprise-level system that supports 18 academic departments across campus with a user-centric service model. Our team, the WFU HPC Team, is comprised of three members—Adam Carlson (Assistant Director of Research Computing), Cody Stevens (Senior HPC UNIX Systems Administrator), and Dr. Sean Anderson (Senior HPC UNIX Systems Administrator)—who are dedicated to advancing research computing and preparing students for the modern STEM workforce.
By leveraging Open OnDemand's uniquely flexible profile functionality, we have transformed a single cluster access point into a dynamic ecosystem of course-specific learning portals. This approach is highly scalable because it leverages built-in OOD features with minimal extra overhead. Instead of provisioning separate, resource-heavy infrastructure or standalone virtual machines for every class, we host everything on a single Open OnDemand instance.
We maintain control over our DNS entries, allowing us to readily create a domain for each course and issue a wildcard SSL certificate with automated DNS verification via AWS Route 53 with no manual intervention. The OnDemand profiles are keyed for each unique domain (e.g., sta379.deac.wfu.edu or phy262.deac.wfu.edu). As a result, the backend infrastructure required to deploy an entirely new portal is almost fully automated.
Standing up a new course dashboard requires configuring just three primary assets:
- A Portal Profile YAML File (e.g., profile-sta379.yml): Hardcodes the specific cluster configurations, custom titles, and environment paths.
- A Custom HTML Widget: A clean HTML file that integrates course descriptions, documentation, video tutorials, syllabus requirements, and direct platform links. This file supports Google Drive integration and basic LMS links.
- Custom Visual Assets: Branded logos tailored and generated via ImageMagick to give each department or class a personalized, institutional feel.
When students log into their dedicated course portal, they are not greeted by a blank Linux terminal. Instead, they see a highly curated dashboard containing only the exact software applications required by their professor. Their class directories are pre-configured and appear in the navbar automatically. For the most part, faculty and students can start using the portal and applications immediately, without any prior training or intervention from the HPC Team.
To shield critical research lines from the sudden bursts of high-density classroom traffic, we maintain an ongoing Slurm reservation composed of 16 compute nodes, 704 CPU cores, and 2.9 TB of memory, with a 24-hour maximum job wall-time. We call this the DEAC Educational Cluster. It guarantees continuous and immediate job allocation, ensuring that student applications flow smoothly during peak grading and lab cycles.
This profile-driven approach has fundamentally altered how professors manage computational coursework. Because each individual course portal automatically funnels assignments into a specific Slurm allocation, faculty can receive granular cluster utilization reports. A quick audit of the Slurm report allows instructors to effortlessly verify lab participation and proactively identify students who may be struggling to complete their assignments.
You can try this approach out for yourself! We have a full tutorial on GitHub.