Extreme-scale readiness program doubles its impact

Ten research projects were successfully granted access to the first Pawsey Centre for Extreme scale Readiness (PaCER) program, establishing Australia’s research platform for extreme scale computing. PaCER aims to ready the research community to make full use of Pawsey’s next era of supercomputing from late 2022.

The PaCER program represents an opportunity for researchers to become exascale-ready, working on new algorithm designs and optimising codes, workflows and data movement pipelines.

Dr Maciej Cytowski, Pawsey’s Head of Scientific Services and co-Chair of the PaCER Committee, describes the PaCER program as: “a long-term partnership between researchers and Pawsey’s supercomputing specialists to achieve extreme performance on next-generation supercomputers.  This will unlock research at previously unavailable scales to benefit not just PaCER participants but eventually the broader research community”.

Researchers will gain early access to supercomputing tools and infrastructure, training and exclusive hackathons focused on HPC performance at scale.

In 2022 they will be the first to exploit the scalability offered by Pawsey’s new 50 petaFLOPS supercomputer Setonix and solve ‘grand challenge’ problems currently unachievable with the current Magnus and Galaxy supercomputers.

The successful projects represent collaborations between ten Australian universities, CSIRO, ICRAR, and internationally with the SKA, US National Laboratories (Argonne National Labs, Ames National Labs, and the Oak Ridge National Labs), University of Toronto, and industry partner General Electric through GE Global Research.

The quality of the applications was such that Pawsey doubled the number of positions available with the support of Pawsey, Capital Refresh and PaCER governance.

Each project will see Pawsey co-fund a doctoral or postdoctoral position and provide Pawsey expertise to work closely with the research team.

The ten successful PaCER projects and their principal investigators are:

  • Dr Giuseppe Barca, “The EXtreme-scale Electronic Structure System (EXESS): predicting the chemistry of nanomaterial interfaces”, in the domain of quantum chemistry
  • Professor Debra Bernhardt, “Towards a molecular level understanding of flow-induced physical and chemical reactions”, in the domains of statistical mechanics and rheology
  • Professor Igor Bray, “Calculation of collisions with molecular targets using the convergent close-coupling method”, in the domain of atomic and molecular physics
  • Professor Melanie Johnston-Hollitt, “Parallel interferometric GPU imaging”, in the domain of radio astronomy
  • Dr Waseem Kamleh, “Emergent phenomena revealed in subatomic matter”, in the domain of nuclear physics
  • Dr Christopher Leonardi, “Massively parallel models of particle suspensions”, in the domains of computational fluid dynamics, geoscience and petroleum
  • Dr Martin Meyer, “Delivery of a next-generation data storage approach to unlock deep SKA and pathfinder observations”, in the domains of radio astronomy and cosmology
  • Professor Richard Sandberg, “Towards exascale simulations for efficient, low-emissions gas turbines”, in the domains of computational fluid dynamics, turbulence and engineering
  • Dr Pat Scott, “Searching for new particles from the attoscale to the exascale with GAMBIT”, in the domain of particle physics.
  • Dr Marcin Sokolowski, “BLINK and you’ll miss it: blazingly fast all-sky radio astronomy pipelines”, in the domain of radio astronomy

Five successful projects were already Magnus users, Pawsey’s current flagship system,  with allocations that represented more than 20% of Magnus total capacity. In contrast, there are two projects entirely new for Pawsey. Three projects are addressing data processing and computational challenges associated with radio astronomy.

Six of the PaCER projects also have an allocation on the National Computational Infrastructure’s supercomputer Gadi. We anticipate that the work done with Pawsey through their PaCER project will also translate to further research efficiencies in their work with NCI and with other supercomputing centres worldwide.

Dr Cytowski points out that a major idea driving PaCER is to get Pawsey deeply involved in software development for exascale computing.

“Eight of these projects are focused on developing their own code for their research problems, and Pawsey will be working closely with them on software development and optimisation.  We aim to build exascale computationally-literate communities around the various scientific domains we are supporting, through both our collaborative project model, and by sharing experience and results with the wider research community,”  said Dr Cytowski.

In addition to the successful projects, the quality and scope of a further eight projects were so high that Pawsey has committed to supporting them through Pawsey Uptake Projects, a collaboration model used by Pawsey supercomputing team in previous years.

The eight Pawsey Uptake Projects will be part of  PaCER training and hackathons, with early access to the same supercomputing tools and infrastructure.  These projects extend the domain areas covered by PaCER to aerodynamics, geodynamics, tectonics, bioinformatics, biomolecular modelling, soil degradation and climate change.

PaCER project teams will work to determine the best ways to scale their research to Setonix.  “Each project has a ‘grand challenge’ – a scalable problem tackling their major scientific question that is not possible to address using current supercomputing resources.”

“Both Setonix and these research teams will be ready for these ‘grand challenges’ in late 2022, demonstrating pathways to what is currently unachievable in Australian research.”

This story was written by Kitty Drok – from Transmogrify

Dr Maciej Cytowski, Pawsey’s Head of Scientific Services and co-Chair of the PaCER Committee