After another year spent adapting to the spikes and lulls in COVID-19 risk levels, ALCF staff are finally seeing a return to normalcy. We can report that thanks to many committed individuals and a robust safety culture, three-quarters of Aurora’s nodes are now in place, with the remaining installation continuing through spring and early summer 2023, and capping off with an upgrade to Sapphire Rapids CPUs with high-bandwidth memory. The deployment of Aurora is the facility’s most important current activity, and we are proud of our progress thus far.
But our activities are not limited to Aurora. The ALCF rolled out two powerful new resources in 2022, with the August deployment of the facility’s most powerful supercomputer to date, Polaris, followed by the December launch of Aurora’s test and development system, Sunspot. Both machines are part of a plan to ready users to do science in the exascale era – the rapidly approaching era of Aurora.
Polaris supports research teams from DOE’s Exascale Computing Project (ECP) and ALCF’s Aurora Early Science Program (ESP) who are working to port and improve the performance of various codes, frameworks, and libraries, while serving as a platform for efforts to couple ALCF resources with large experimental facilities. Sunspot also supports ECP and ESP teams doing porting, scaling, and performance optimization research. ALCF’s ever-growing AI Testbed is now one of the nation’s most diverse collections of AI technologies dedicated to open science. Two of its systems are now available to the research community, with more to come next year.
Also, with Polaris, we and other computing colleagues at Argonne are continuing the laboratory’s more than 75 years of history of integrating research infrastructure with computational resources. In 2022, staff at Argonne’s Advanced Photon Source (APS), Globus project, and ALCF worked on coupling our resources with those at the APS, demonstrating a complete end-to-end (no “humans in the loop”) acquisition of data, computational analysis, and sharing of the results. This example of a strong partnership between domain scientists, ALCF, and the Globus project team represents a new way of driving breakthrough discoveries, and we look forward to building on these successes in the years ahead.
In 2022, we made significant strides in our commitment to education and workforce development by offering a range of information-rich learning experiences and training events. ALCF’s 2022 AI for Science series alone attracted more than 200 attendees from 90 universities, providing them with hands-on experience using AI for scientific research. We also expanded the reach of our educational offerings by adding sessions from ALCF’s 2022 Aurora-related webinar program to our YouTube channel, making them accessible to a wider audience.
While we are proud of these achievements, we recognize that they are just the first steps toward ensuring that we have a computationally savvy workforce. In the years ahead, we will continue to develop new and innovative programs that engage students and researchers at all levels, from kindergarten through postdoc. By working closely with our partners in academia, industry, and government, we will help to cultivate the next generation of computational scientists and engineers, empowering them to drive scientific discoveries and technological innovations that will shape the future of our world.
We can also report that in 2022, ALCF’s stalwart production system, Theta, delivered over 20 million node-hours to INCITE projects, more than 7 million node-hours to ALCC projects, and over 5 million node-hours to smaller, vital discretionary projects that often serve to prepare new users for larger awards.
Theta and Polaris have together advanced studies ranging from research to develop first-principles simulations of black hole accretion flows and coronae, to ongoing work to advance our understanding of supernova explosions, to designing the next-generation of batteries, to performing extreme-scale simulations for advanced seismic ground motion and hazard modeling. And, notably, ALCF resources supported the Argonne-led research team awarded the ACM Gordon Bell Special Prize for HPC-based COVID-19 Research for their efforts to create the first genome-scale language models (GenSLMs) for understanding the evolution of SARS-CoV-2 and other pandemic-causing viruses.
In closing, we are deeply committed to our mission of enabling groundbreaking scientific discoveries through our unique resources, capabilities, and deep research expertise. We remain steadfast in our dedication to expanding our educational initiatives to reach a broader audience and are excited about the possibilities. As we look to the future, we recognize the critical importance of data in driving scientific progress. Whether it is computed or collected from sensors, data must be opened and shared to enable collaboration and breakthrough discoveries.
With your continued support, we are confident that we can continue to advance the frontiers of scientific research and support the next generation of scientific discoveries. We want to take this opportunity to express our heartfelt thanks to our dedicated staff, vendors, and program office, without whom our achievements in 2022 would not have been possible. We look forward to sharing more details about our accomplishments in this year’s Annual Report.