Challenge
An aging HPC cluster that could not keep pace with ML and AI research, serving the whole campus on a strict public budget that made an open-source software stack a firm requirement.
Solution
A ground-up rebuild: a new eight-GPU cluster, a 100-gigabit storage network, a fully open-source HPC stack, command-line and web-based workflows, a standardized environment and performance tuning.
Outcome
Roughly 2x faster job performance, validated side by side against the legacy cluster, with no software licensing fees and a follow-on order to extend the cluster.
Service
Project Services
Date
October 28, 2026
Pages
3
Industries
Higher education