Tailoring HPC for medical research: lessons from the Melbourne Parkville Precinct

HPC as a research tool has existed for decades, but HPC uptake in Medical Research Institutes (MRIs) is relatively new, driven by the proliferation of genomics and bioinformatics; the advent of CryoEM and imaging technologies; and more recently, AI technologies like Alphafold. This uptake is still increasing as more medical researchers are integrating HPC into their research, prompting MRIs to establish their own on-premises HPC. The narrower target audience at MRIs allows for a tailored HPC solution, but the nature of the research and the research software presents its own challenges. This talk presents perspectives from institutes in Melbourne s renowned Parkville medical research precinct (WEHI, MCRI, Peter MacCallum Cancer Centre, as well as Melbourne University), regarding the adaptation of the HPC concept for MRIs. The talk covers challenges and solutions (if any) for the following areas: 1.Key research technologies and software driving the hardware and filesystem decisions and HPC scheduler policies decisions 2.Researcher profiles and implications for cluster usage, including the necessity of interactive jobs, facilitated by Open OnDemand and other GUI platforms 3.Data management considerations, including HPC filesystem considerations, sensitive data handling, and sharing. 4.Important learnings taken from more established HPC facilities. 5.Finally, ongoing challenges will be presented.