NUCLEAR RISKSNuclear Agency Cannot Continue With “Business as Usual” in the Shifting Supercomputing Landscape: Report

The National Nuclear Security Administration (NNSA) needs to fundamentally rethink the strategy for its next generation of high-performance computing, says a new report from the National Academies of Sciences, Engineering, and Medicine. In recent decades, the NNSA had a clear technical path forward in advanced computing and was able to leverage the growth path of the computing industry. However, these conditions no longer hold, the report warns; a simple extension of past strategies will be insufficient if it is to be successful in acquiring post-exascale computers. 

A core mission of the NNSA is to ensure that the U.S. maintains a safe, secure, and reliable nuclear stockpile through the application of science, technology, engineering, and manufacturing. Advanced computing capabilities allow NNSA scientists to conduct sophisticated modeling and simulations, serving in the design, redesign, engineering qualification, surveillance, maintenance, and certification of the nation’s aging nuclear stockpile. These are key parts of responsible, science-based stockpile stewardship, especially since 1992, when the U.S. ceased underground nuclear explosives testing.  

The report finds that demands for advanced computing at NNSA will continue to grow and will exceed the capabilities of planned upgrades, which include the installation this year of the second exascale computing system in the U.S.“Exascale” refers to a computer that performs a quintillion (10¹⁸ or 1,000,000,000,000,000,000) floating-point operations per second. As it looks beyond exascale, NNSA needs to focus on building computers that can reduce the time needed for its largest calculations from months to days. Ensuring that future weapons systems will function during atmospheric reentry, assessing the effects of aging plutonium in the stockpile, and assessing margins and uncertainties for current and future weapons systems are all challenges for the NNSA that will require post-exascale computing.  

“The NNSA has had a really successful run over the last 30 years with a combination of high-end computing facilities and expertise in computational science that make its labs a critical national resource,” said Kathy Yelick, chair of the committee that wrote the report, and vice chancellor for research and Robert S. Pepper Distinguished Professor of Electrical Engineering and Computer Sciences at the University of California, Berkeley. “But between the rapidly evolving geopolitical situation ―which reinforces the need for computing leadership as an element of deterrence ―and challenges in computing technology, the NNSA’s approach must be reimagined for future success.” 

The NNSA should develop and pursue new and aggressive design, acquisition, and deployment strategies, the report says, as the agency’s computing needs have increasingly diverged from the interests and products of the commercial market on which it relies. The NNSA must find ways to either leverage the burgeoning hyperscaler (large cloud provider) and artificial intelligence hardware markets or work with other government partners and vendors to design computing systems suitable for NNSA mission problems. The report warns of other issues presenting technical and economic risks for the NNSA, including reliance on foreign semiconductor fabricators, and the fact that all its exascale systems are being integrated by a single company. 

Attracting and retaining top talent is a significant challenge for the NNSA’s labs, the report says. Private industry can offer compelling benefits, including higher salaries, greater resources, and more flexible work environments. Also, recruiting foreign talent is complicated by security concerns, which limits NNSA hiring options. The report urges the NNSA to address its workforce challenge through development of a national strategy involving partnership across agencies and academia and an international strategy that creates a welcoming environment for diverse talent. 

Among the report’s other findings and recommendations: 

·  The NNSA should develop a road map for meeting its computing needs while accounting for the technology and market challenges. 

·  The NNSA should foster and sustain high-risk, high-reward research in applied mathematics, computer science, and computational science to shape its road map and ensure needed future intellectual leadership. 

·  The NNSA should expand research in artificial intelligence methods, and evaluate and improve these methods in the performance of its mission; these are likely to complement rather than replace simulation. 

·  The NNSA should continue to investigate and track innovations in quantum computing, but algorithm and system breakthroughs are required to be useful in multiphysics stockpile modeling. 

The study, undertaken by the Committee on Post-Exascale Computing for the National Nuclear Security Administration, was sponsored by the National Nuclear Security Administration.