National Energy Research Scientific Computing Center (NERSC) director Sudip Dosanjh kicked off NERSC’s 50th anniversary year in front of a crowd of NERSC users, staff, and fans in the U.S. Department of Energy (DOE) booth on Nov. 14 at the International Conference for High Performance Computing, Networking, Storage, and Analysis 2023 (SC23). Beginning with an introduction by Advanced Scientific Computing Research (ASCR) Facilities Division Director Ben Brown and a video exploring the center’s roots and its role in HPC going forward, Dosanjh launched the coming year of celebrations with his featured talk, “The Next 50 Years: How NERSC is Evolving to Support the Changing Mission Space and Technology Landscape.”
Throughout 2024, NERSC will be highlighting 50 years of scientific achievement and shining a spotlight on the ways the center is moving into the next half-century.
Other conference highlights were spread broadly across the Computing Sciences Area (CSA). Former CSA associate lab director and former NERSC director Kathy Yelick brought her perspective on the past, present, and future of HPC to the SC23 week by participating in Monday’s SC23 plenary panel, “I Am HPC,” highlighting the people of HPC and analyzing the social impact of HPC through scientific and technical achievements and resources.
Also on Monday, in a meeting hosted by Berkeley Lab, nearly 200 researchers from eight national laboratories gathered to share their latest research efforts with several ASCR program managers. During the two-hour event, researchers had the opportunity to present “lightning talks” about their work to the panel of program managers, followed by an interactive panel discussion. The summit provided attendees the opportunity to network with peers and program managers and discuss opportunities for collaboration across the ASCR ecosystem. The event was “highly successful,” noted Stefan Wild, director of the Applied Mathematics and Computational Research (AMCR) division director and one of the organizers, and will likely be held annually going forward.
Throughout the week, among a record attendance of more than 14,000, CSA researchers, scientists, and engineers represented Berkeley Lab through a range of programming. Their participation included 28 workshops, 11 Birds of a Feather presentations, six posters, five tutorials, four panels, and four papers, in addition to presentations at the DOE booth comprising ten technical demonstrations, four roundtables, and two talks, including Dosanjh’s 50th anniversary kickoff.
Among those representing the research divisions, John Shalf participated in the panel “Chiplet Ecosystem in High Performance Computing, AI/ML, and Data Acceleration,” moderated by Bapi Vinnakota. The panel discussed the possibility of an ecosystem or marketplace of chiplets available for system developers to use to build next-generation devices and weighed the pros and cons of off-the-shelf chiplets compared with custom-designed chiplets.
Additionally, Applied Mathematics and Computational Research scientist Anastasiia Butko and former Berkeley Lab and ESnet scientist Mariam Kiran drew a crowd to the DOE booth with their technical demonstration “5G on the Show Floor,” while Scientific Data Division scientists Jean Luca Bez, John Wu, and Suren Byna and research intern Hammad Ather presented on “Drishti: Where is the I/O Bottleneck?”
A number of NERSC staff and users also presented across a range of workshops, panels, and demonstrations, including a well-attended technical demo by Sam Wellborn, Bjoern Enders, Peter Ercius, Chris Harris, and Debbie Bard on the live-streaming of large-electron microscope data to NERSC, and a related roundtable with Bard, Laurie Stephey, and David Schissel of General Atomics – just two of many sessions highlighting Berkeley Lab’s leadership in establishing the superfacility model and IRI for scientific research.
A paper with ties to Berkeley Lab and the DOE Exascale Computing Project was a finalist for the SC23 Best Paper award; “Frontier: Exploring Exascale” considered the Frontier supercomputer at Oak Ridge National Laboratory (ORNL) and its architecture and deployment in the context of DOE’s 15-year journey toward exascale computing. Axel Huebl and Jean-Luc Vay of the Berkeley Lab Applied Physics and Accelerator Technologies Division and Andrew Myers of the Berkeley Lab Center for Computational Sciences and Engineering contributed to the paper.
Oscar Antepara, Sam Williams, and coauthors were also awarded Best Paper at the Tenth Workshop on Accelerator Programming and Directives (WACCPD’23) for their paper “Performance-Portable GPU Acceleration of the EFIT Tokamak Plasma Equilibrium Reconstruction Code.”
Representing the Energy Sciences Network (ESnet) was Scientific Networking division director and ESnet executive director Inder Monga, who joined Brown and ORNL’s Nagi Rao for a lively Birds of a Feather talk on continuum computing organized by Neena Imam from Southern Methodist University. Continuum computing — which Imam defined as “executing complex workflows over ecosystems of networked computing systems and instruments with distributed, coordinated computations orchestrated across the edge, cloud, datacenter, and HPC computing platforms” — is aligned with DOE’s new Integrated Research Infrastructure (IRI) program, Brown explained, emphasizing that software will also be a key part of this infrastructure, as will ESnet, binding it all together.
In addition, several ESnet staff held leadership roles among 200-plus volunteers who built SCinet, the Wide Area Network connecting the attendees to each other — and to scientific research sites around the globe, for demanding technical demonstrations. At 6.71 terabits per second, SCinet was temporarily the world’s fastest network and battle-tested several new technologies, including the new Wi-Fi 6E standard.
At the end of the week, two projects with NERSC connections were honored with the sought-after ACM Gordon Bell Prize and the first-ever ACM Gordon Bell Prize for Climate Modeling. A collaboration hailing from the University of Michigan, the Indian Institute of Science, and Oak Ridge National Laboratory took home the much-coveted Gordon Bell Prize for outstanding achievement in high-performance computing for their project, “Large-scale Materials Modeling at Quantum Accuracy: Ab Initio Simulations of Quasicrystals and Interacting Extended Defects in Metallic Alloys.” Separately, the team used the Perlmutter supercomputer at NERSC to study the stability of quasicrystals in an ytterbium-cadmium alloy.
Additionally, the new Gordon Bell Prize for Climate Modeling went to a multi-lab collaboration using Perlmutter as part of its work with the Energy Exascale Earth System Model (E3SM) to support its project, “The Simple Cloud-Resolving E3SM Atmosphere Model (SCREAM) Running on the Frontier Exascale System.”
About Computing Sciences at Berkeley Lab
High performance computing plays a critical role in scientific discovery. Researchers increasingly rely on advances in computer science, mathematics, computational science, data science, and large-scale computing and networking to increase our understanding of ourselves, our planet, and our universe. Berkeley Lab's Computing Sciences Area researches, develops, and deploys new foundations, tools, and technologies to meet these needs and to advance research across a broad range of scientific disciplines.