THE Department of Energy’s Accelerated Strategic Computing Initiative—now the Advanced Simulation and Computing (ASC) Program—has secured an enduring position in Lawrence Livermore’s computing legacy through its advancement of high-performance computing (HPC) hardware, software, and application development in support of science-based stockpile stewardship. The complementary unclassified Multiprogrammatic and Institutional Computing (M&IC) Program, developed by Livermore soon after, helped ensure HPC was available to all Livermore scientists, enhancing the vitality of the institution’s science and technology and helping retain a top-tier scientific workforce.
Today, nearly two decades after it started, M&IC allocates time on unclassified systems through three channels: the Laboratory Directed Research and Development Program, the discovery-focused Computing Grand Challenge Program, and strategic capability banks. The Computing Grand Challenge Program, whose projects consume the largest portion of M&IC resources, was launched in 2005 to provide unclassified capability computing that would allow researchers to exploit the significant advantages of working at scale. As discussed in Science on a Grand Scale, the Computing Grand Challenge Program grants scientists with big ideas the computing resources they need to tackle otherwise insurmountable mission-relevant scientific challenges.
Classified and unclassified computing have a synergistic relationship. Advances in algorithm development or modeling made through Grand Challenges and other unclassified research initiatives often benefit classified research efforts, while supercomputing infrastructure and expertise developed for classified research enhances unclassified resources. For instance, Vulcan, the most powerful system available to Grand Challenge winners, is a smaller version of Livermore’s leading ASC machine, Sequoia, the world’s third most powerful supercomputer, which Livermore computer scientists helped design.
Livermore’s world-class HPC ecosystem—its hardware, software, and computer support staff—empowers scientific discovery and technology development on both the classified and unclassified fronts in support of our institution’s vital missions for the nation. Maintaining the necessary competitive edge in the rapidly evolving world of HPC requires that we work on three timescales simultaneously: delivering capabilities today, innovating for and investing in tomorrow, and envisioning what will come the day after tomorrow, that is, what is next on the horizon.
We are currently addressing the “tomorrow” trajectory in several exciting ways. One is by modernizing our computing infrastructure. In May, the Laboratory broke ground on a modular and sustainable supercomputing facility to help meet the growing need for unclassified computing at Livermore and to accommodate a variety of HPC architectures. Another is by preparing existing and new applications to run on Sierra, Livermore’s next ASC advanced technology system. Sierra is part of the Collaboration of Oak Ridge, Argonne, and Livermore (CORAL). A procurement collaboration between the three national laboratories and industry partners, CORAL seeks to develop and deliver next-generation petaflop-scale (1015 floating-point operations per second) systems to the three laboratories in 2017 or 2018. As with Sequoia, Sierra will be made available for marquee unclassified calculations before it enters into classified operations, to identify and resolve any security and stability issues while supporting scientific discovery.
The “day after tomorrow” scenario is computing at the exascale—speeds of 1018 flops and memory of 1018 bytes. Our computer scientists have long been planning and preparing for extreme-scale computing, but a new multi-agency government initiative is bringing it one step closer to realization. In July, President Obama established the National Strategic Computing Initiative, which aims to create, before 2025, the world’s first exascale computer. This machine will be more than 30 times faster than today’s fastest systems. Livermore and other national laboratories will play a key role in the design and development of this new class of computers. Such initiatives will help keep the United States at the forefront of HPC capabilities and enable institutions such as Livermore to continue pushing scientific and computational boundaries.
Modern scientific advances rely on a robust and dynamic computing ecosystem, which Lawrence Livermore maintains through programs such as ASC and M&IC. Strategically deploying these resources through the Computing Grand Challenge Program and other efforts supports our national security missions and gives researchers some of the tools they need to take on big scientific and engineering challenges.