Managing Computing Resources with Slurm for Multiple Courses Usage

In this work, we present a scalable and fair resource management solution to support multiple courses. We replace the previous direct-access system with Slurm, a widely used workload manager, to improve fairness, utilization, and manageability. The deployment is automated using IaC tools, including PXE and Ansible, enabling us to provision a fully functional Slurm cluster from operating system in minutes. To support diverse software requirements across courses, we integrate package managers such as Guix and Conda, ensuring consistent and reproducible runtime environments. We further employ Slurm’s QOS features to enforce per-user resource limits and multifactor scheduling policy. Our system has been deployed in the Spring 2025 semester to support the Parallel Programming course, with 49 active users and approximately 92,000 job submissions. Despite the high demand, our analysis shows that fewer than $\mathbf{1 0}$ nodes were sufficient to handle the workload, demonstrating the effectiveness of proposed architecture.