Expanded Supercomputing Maximizes Scientific Discovery
HIGH-PERFORMANCE computing has been a part of the Laboratory’s genetic code since it was founded 54 years ago by physicists Ernest O. Lawrence and Edward Teller. Although computer simulation was in its infancy at the time, our Laboratory leaders recognized its potential to accelerate scientific discovery and engineering underpinning Livermore’s national security missions. Our ability to use computers for simulating the physical world has evolved to the point where today three-dimensional simulation is an integrating component of the scientific method of observation, theory, and experiment. Livermore continues to develop increasingly more powerful, scientific computing tools and techniques that enable scientific discovery and the fulfillment of Laboratory missions.
Strategic leadership combined with a willingness to explore new directions in scientific computing over the years have established the Laboratory as one of the world’s premier institutions for computer and simulation science. By leveraging the capabilities of the Advanced Simulation and Computing (ASC) Program, an important component of stockpile stewardship, our Multiprogrammatic and Institutional Computing (M&IC) Program has extended leading-edge simulation on scalable platforms to unclassified science in the national interest. M&IC works to balance investments in an effort to provide cost-effective capability platforms and meet the ever-growing demand for high-performance computing capacity from the broad cross section of Laboratory researchers.
Livermore’s Thunder supercomputer is a shining example of our success with this high-performance computing strategy. Brought on line in 2004, Thunder is a 23-trillion-floating-point-operations-per-second (teraflops) cluster of 4,096 processors. When it debuted, Thunder was the world’s second fastest supercomputer on the Top500 list according to the industry standard LINPACK benchmark. Two years later, Thunder still holds a respectable ranking of 14 in the highly competitive and rapidly developing supercomputing environment. Systems such as Thunder represent a shift from traditional parallel computers to Linux cluster technology, which offers increased power at a lower cost. M&IC’s versatile systems are a product of the Laboratory’s consistent investment, and the result is a powerful unclassified computing resource that is being used to push the limits of computing and its application to simulation science.
Over the last year, the M&IC Program has succeeded in using Thunder to run “Grand Challenge” projects—a select number of unclassified, mission-relevant projects that are allocated large banks of computing cycles to aggressively tackle problems for breakthrough science. These projects, as described in Thunder’s Power Delivers Breakthrough Science, range from research in climate change to the interface of water and vapor to nanofluidics to dislocation dynamics to protein folding. All of these projects involve collaborations with other leading research institutions such as the Scripps Institution
of Oceanography, the National Center for Atmospheric Research, the IBM Almaden Research Center, the University of California campuses at San Diego, Berkeley, Davis, and Irvine, the University of Zurich, and the Polytechnic of Torino. Our capabilities and knowledge make us a desirable partner for collaborations that can accelerate scientific discovery to benefit both the Laboratory and the larger scientific community.
While Livermore’s high-end classified systems BlueGene/L and ASC Purple attract the spotlight because of their respective number one and three world rankings on the LINPACK benchmark, systems such as Thunder are sometimes used to lay the scientific computing groundwork that is later transferred to the more powerful machines. Thunder is also paving the way for the next generation of supercomputers recently acquired as part of the Peloton procurement, which is bringing to the Laboratory an additional three clusters named Atlas, Zeus, and Rhea for a total of 77 teraflops. Atlas will take on M&IC’s heavy lifting as the 44-teraflops capability machine for Grand Challenge projects starting in January.
The environment created to develop and sustain the ASC supercomputers for simulating nuclear weapons performance is also fostering the development of increasingly powerful unclassified systems for conducting complementary science. By ensuring that classified and unclassified computing work in complement, we’ve greatly expanded the high-performance computing resources available to Laboratory researchers in a broad range of disciplines. This symbiosis is another facet of the multidisciplinary team science that is a founding principle of Livermore. Our continued success is crucial not only to our national security missions and basic science but also to our ability to attract the top scientific talent needed to keep our Laboratory on the cutting edge.