Ogston S1, Crosby S1, Giugni S1
1The University of Melbourne, Melbourne, Australia
Biography:
TBD
Abstract:
Over the past decade, research computing platforms have evolved to meet rapid growth in data, compute demand, and AI-driven research. At the University of Melbourne, Research Computing Services (RCS) has continuously evolved its HPC, cloud, and data platforms to support increasingly complex and diverse research needs.
This presentation reflects on what worked—and which early assumptions did not scale—in evolving these platforms. Key challenges included integrating heterogeneous infrastructure, responding to rapidly changing workloads and user demand, and balancing flexibility with standardisation.
Through iterative redesign, RCS has shifted from infrastructure-centric delivery to a more integrated, service-oriented model, improving interoperability, supporting GPU-enabled workloads, and strengthening data lifecycle management and governance.
We share practical lessons across platform architecture, data and compute integration, and service design, highlighting trade-offs that shaped current capabilities.
These lessons provide a grounded foundation for evolving research platforms to support AI-enabled, data-intensive research at scale, emphasising integration, adaptability, and learning from both success and failure.
