Pascal Elahi could always see that HPC would revolutionise scientific research by enabling experiments that wouldn’t be possible in the physical world. “I’ve always been fascinated by the history of the cosmos, and how physics governs its evolution,” he explains. “But that’s difficult to explore experimentally. The idea of using HPC systems as a virtual laboratory for building test universes to explore the physics that govern our own Universe was very appealing, and what drew me to the field of computational astrophysics”.
Pascal is now a Supercomputing Applications Specialist at Pawsey, using his science background and understanding of how researchers use HPC to help Pawsey users run their software on large supercomputing systems efficiently. If your model system spans the Universe, you’ve got to run it efficiently, no matter how big the computer is.
His fluency in programming languages like C++, CMAKE, Python, and GPU programming, and communication and training skills are being used to help others gain maximum benefit from their HPC use. Pascal particularly focuses on the radio astronomy and bioinformatics communities, and is now deeply embedded in their activities at Pawsey.
“Using HPC systems is not always a simple task, and tailoring software to make the most efficient use of Pawsey’s new systems requires expertise in software development and computing, along with an understanding of the research being attempted,” says Pascal. “The key goal of my work is to enable new science by allowing researchers to perform new, technically-challenging simulations and analysis – their critical computations – on our supercomputing infrastructure.”
Pascal particularly appreciates the diversity of science it enables, from small research projects to mega-science projects like the SKA. As an example, Pascal has been helping the radio astronomy community using the MWA to migrate onto the new compute cluster Garrawarla, procured as part of the Capital Refresh, and improve their analysis pipeline to make the most of the latest-generation Central Processing Unit (CPU) and Graphics Processing Unit (GPU) processing now available. He also helped the ASKAP (Australian Square Kilometre Array Pathfinder) operations team commission their new compute cluster, designed to ingest and process data coming directly from the telescope. Extensive collaboration with the ASKAP team was required to get the new system running the complex software needed to process ASKAP data on schedule.
“HPC is moving towards more specialised accelerators making up a significant portion of the computing power provided, specifically GPUs as we’re seeing here at Pawsey, and even eventually QPUs (Quantum Processing Units),” notes Pascal. “So we will always need software development and computing specialists to enable researchers to make the best use of the computational power available.” And he’s not complaining – the diversity of his working day, from meeting with researchers, training users, documenting computational processes and troubleshooting problems – is one of his favourite parts of working at Pawsey.
Pascal’s future at Pawsey looks to be even more varied: “Our research users themselves are now joining us in pushing to transition to more ‘green’ computing, such as our new more energy-efficient GPU clusters. The carbon footprint of computing is significant, but by staying deeply embedded with the research communities we support, we’re well positioned to keep Australian researchers at the forefront of energy-efficient computational ability as it continues to improve.”