"The Spartan HPC Story: From Small Scale Experimental to Top500 and Beyond"

Faced with a constrained but an increased need for a general purpose High Performance Computer (HPC) the University of Melbourne s , Spartan , was first made available for researcher use in 2016. A needs evaluation led to the implementation of highly innovative architecture at the time, making use of RoCE and a Cloud-HPC hybrid partition structure. Whilst initially a small system, Spartan attracted a lot of attention with its design with presentations in New Zealand, Australia (eResearch), several European HPC centres (including CERN), and the OpenStack Summit In Barcelona. These early successes led Spartan to receive a substantial grant for a GPU partition for a consortium of Victorian universities that pushed the system into the same metrics as a Top500 system, although it was not until November 2023, that certification was sought and achieved (and that was just a single partition). From a small innovative and cost-effective system designed for general needs, to a world class system, a retrospective of Spartan s history provides insight for information systems management and operational implementation, covering architecture, hardware choices, user engagement, and training. Further, with future portents under consideration, this presentation will outline a roadmap on Spartan s probable future as it heads toward a ten years of operation which, like the fabled Ship of Theseus, engages in a process of development and regeneration.