Second place on supercomputers not good enough
When the GE Global Research team wanted to develop quieter, more fuel-efficient jet engines, they faced a major problem: They needed to know how changes in the plane's flap positions would affect the engine's performance -- a tough question to answer, given all the possible variables of engine wakes, wind speeds and weather conditions.
So GE found a next-generation solution: The company turned to Mira, the new supercomputer at Argonne National Laboratory, to create advanced, realistic jet engine simulations that will be used to shape their future engine designs. Already, it appears that the new design could help airlines trim 1 percent from current fuel usage, roughly equivalent to 6.7 million barrels of oil every year.
Across the country, researchers and private companies are using supercomputers built by the U.S. Department of Energy to conduct cutting-edge research and solve tough problems on a scale never before seen.
Mira, DOE's latest supercomputer, is now ranked as the world's fifth-fastest, with the capability to perform 10 quadrillion calculations per second. Mira's power allows researchers to tackle difficult challenges, such as optimizing Illinois' power grid, and gives businesses like GE a chance to bring their R&D to the next level. Access to Mira and other top-tier supercomputers gives American science and industry a leg up on competition around the world. However, as the next generation of supercomputers begins to emerge, our country's ability to maintain that advantage is far from assured.
Currently, our nation is home to 252 of the 500 fastest computer systems worldwide. But China, Japan, India, Russia and the European Union are making investments that challenge U.S. leadership in this vital field. In June, the new Tianhe-2 supercomputer at the National University of Defense Technology in China vaulted to the top of the world rankings. At its peak speed, this new Chinese computer is almost four times faster than its nearest rival.
South Korea, Japan and China are doubling their investments in supercomputing because it saves time, money and energy, strengthening a country's economy and national security. The United States can't afford to fall behind.
To keep America in the forefront of supercomputing, the DOE and our national laboratories are looking at ways to develop and build new exascale computing systems. Those next-generation machines will be 1,000 times faster than even petascale computers like Mira. To support this effort, the U.S. Senate's 2014 Energy Appropriations bill would provide $150 million for the Exascale Computing Initiative at DOE. The Senate also is considering the ExaSCALE Computing Leadership Act of 2013, which would create public-private partnerships to research and develop next-generation exascale computing systems by addressing the major technological barriers in power, hardware, reliability, memory and software.
As we look ahead, it's clear that the road to exascale computing will require substantial public investment. But that expense must be measured against the enormous long-term costs our nation will face if we abandon our quest for leadership in high-performance computing. The country that wins the race to achieve exascale computing capabilities will gain significant intellectual, technological and economic advantages over other nations.
By investing in high-value computer R&D, our federal government will give the United States a real shot at crossing that finish line first, giving American companies an invaluable tool to create good, lasting jobs in tomorrow's growth industries. Ultimately, Mira and other supercomputers are more than world-leading models of computing excellence and efficiency; they really are time machines, propelling American science and industry into a new and even brighter future.
• Illinois Sen. Dick Durbin is a lead sponsor of the ExaSCALE Computing Leadership Act of 2013. Dr. Eric Isaacs is director of Argonne National Laboratory.