On the Trail of the Pedaflop
May 2006A visionary commitment to infrastructure helps Oak Ridge National Laboratory put the “super” into supercomputing
Fortune 500 company Proctor & Gamble produces Pringles brand potato chips, those uniquely American creations with the double elliptical shape, at a large manufacturing plant in Jackson, Tenn. Supplying all of North America, the plant produces hundreds of millions of chips per day utilizing a highly technical and extremely fast manufacturing process. During it, chips fly through the air at a high rate of speed, are dried and collected in a can, perfectly stacked, unbroken. How did P&G develop such a process? Answer: via the company’s substantial on-site engineering operation and the aid of supercomputing.
Obviously, making potato chips is a more mundane use of supercomputing than for instance using simulation to improve the accuracy of weather forecasting, which could avert great human loss due to a tsunami or hurricane. But the Pringles example does exhibit supercomputing’s role in innovation and its economic consequences in the market. It bodes well for Tennessee, which is on the precipice of becoming the dominant global force in computational science.
THE GREATEST RACE
According to Moore’s Law, a theory predicting technological progress that was coined by Intel Corp. founder Gordon Moore in 1965, computers double in speed every 18 months. Precise or imprecise as a measurement, the point is clear. Today’s computers are more powerful than all earlier computers put together. It’s a rate of growth that also holds true for supercomputers. Currently, the world’s fastest public use, non-classified supercomputer is Japan’s Earth Simulator, which runs at about 45 trillion calculations per second, or 45 teraflops in science speak. Previous studies have theorized that to match what Earth Simulator achieves each second would require every living human being working on a calculator around the clock for approximately a decade.
But as Moore’s Law infers, computing doesn’t stand still. And by 2008, the enhancements now being made to the high-end supercomputers at Oak Ridge National Laboratory will culminate in the unveiling of a system capable of performing 1,000 trillion calculations per second. (For those keeping score at home, 1,000 teraflops equals one pedaflop.) When this happens, it will make East Tennessee home to the fastest supercomputer for open research on Earth. Take that, world.
ORNL is already garnering worldwide attention this summer for the launch of the Spallation Neutron Source, the world’s largest public science project. (See cover story in the May edition of Business Tennessee.) That federal decision makers selected Oak Ridge as the site for that project is a boon to all of Tennessee from both a scientific discovery and an economic development perspective. But the same could be said of the supercomputing project. And, as was the case with SNS, ORNL beat out several other national labs to land it. To do so, ORNL engaged in an act of bold speculative real estate that would make even the most visionary commercial Realtor proud.
VIRTUAL WORLD
But first, what’s the big deal about supercomputing? A key benefit of computational research is its ability to yield discoveries through simulation. Using supercomputers, scientists can simulate physical, chemical and biological processes that underlie the behavior of natural and engineered systems in scientific disciplines ranging from life sciences to fusion energy to astrophysics, nuclear energy research, petroleum exploration and nanoscience. Scientists can also carry out virtual experiments that would be too costly, too controversial or too dangerous to conduct employing traditional methods. As an example, scientists wouldn’t want to physically create global climate change in order to study its effects.
Many industries, including aerospace, pharmaceutical, chemical and process industries (like Pringles makers P&G), among others, also rely on high-performance computing for competitive and strategic reasons. For instance, the automotive industry has used simulation to do crash testing on vehicles still under design, minimizing costly post-production failure in field-testing. It’s one reason the automotive design cycle has shrunk from a three- to four-year process to one that requires 18 months to two years to complete. Virtual prototyping both saves money and gets product to market faster.
Sounds important, yes? No wonder ORNL operators so passionately coveted the opportunity to house the world’s fastest supercomputer. But they also coveted it because the mission of computational science meshes so well with the work planned at the Spallation Neutron Source. Once fully operational, ORNL’s supercomputer system will be able to compute phenomena at the same resolution SNS is able to measure it. Said another way, SNS investigates the behavior of material at a much finer scale than ever contemplated before while the enhanced supercomputers will be able to build up material atom by atom in simulation more quickly and powerfully than any other device on the planet. The confluence of those two capabilities could yield striking scientific discoveries in the years ahead.
Becoming the premiere global destination for supercomputing also dovetails nicely with the Laboratory’s thrust as a “nano, info, bio” Laboratory. The value of supercomputers is also high in the biological sciences. Earlier ORNL supercomputers were used for sequencing Chromosomes 5, 16 and 19 of the human genome. ORNL’s efforts as a major player in current genomics research will receive a great boost from its computational foundation. Building the computer infrastructure was part of a broad goal of moving toward these three interlocking research programs.
Now, how did ORNL land the project? Essentially by building a mousetrap. And the vision and risk-taking the lab’s operators exhibited in doing so could soon create a steady stream of Fortune 500 involvement in Tennessee that will bolster the whole state.
SETTING THE SPRING
Housing the world’s fastest supercomputer requires a building with special attributes. UT-Battelle, the lab’s operators, built that building, a nearly $100 million investment, on spec. (By the time the computers are in place, the total project will be valued north of half a billion dollars.) Amenities include an acre-sized raised floor that allows for easy repair and upgrade access, as well as facilitates proper cooling of the computers. Water for cooling is supplied by a 24-inch dedicated water pipe hitched directly to the city of Oak Ridge water supply. And ORNL made an investment in an unusually high amount of network capability (rivaling the world’s best) permitting global access to on location research.
ORNL Associate Laboratory Director for Computing and Computational Sciences, Dr. Thomas Zacharia, gives credit to ORNL’s operators, UT-Battelle, and specifically to former director Bill Madia and current director Jeff Wadsworth, for taking the risk to build the 170,000-square foot space.
“Unequivocally, the risk was taken by UT-Battelle as contractor for the lab,” Zacharia says. “I certainly asked. But the tough decision was made by UT-Battelle with the expectation that the nation needed computing of this scale.”
Regarding the role that having the facility in place played in the Department of Energy’s 2004 decision to grant ORNL supreme supercomputer status, Zacharia says flatly, “If we didn’t have it, I don’t think we would have won the project.”
Other factors contributed to the choice of ORNL. The lab obviously boasts world-class scientific talent; it had the support of the state of Tennessee and University of Tennessee; and it benefited from the nearness of a willing and able energy source in the Tennessee Valley Authority (TVA) to provide the necessary power (eventually 25 megawatts) to power the giant computer. In addition, Department of Energy operatives no doubt took under consideration the advantage of other important scientific work like SNS going on concurrently at ORNL, creating a perfect intellectual environment.
At the time of the announcement, Secretary of Energy Spencer Abraham was quoted stating that the new facility would “serve to revitalize the U.S. effort in high-end computing,” adding “it is no exaggeration to say that this machine will give both the U.S. scientific community and industrial sector a significant competitive advantage over the rest of the world.”
RIDING A WAVE
Reasserting supercomputing dominance is a boost to the U.S. scientific community. But what about the ripple effects of ORNL’s newfound supercomputing status on Tennessee’s business community and the state economy? Will there be any?
President George Bush recently announced his American Competitiveness Initiative, which commits $5.9 billion in fiscal year 2007 to increase stateside investments in R&D, math and science education and entrepreneurship. Bush’s plan specifically points out the role of supercomputing in achieving global competitiveness. For instance, Bush’s push for advances in hydrogen power research could conceivably rely heavily on ORNL and its computational capabilities to speed alternative fuel sources toward commercial viability. ORNL can certainly expect to devour a substantial piece of that new pie.
And don’t forget the research needs of Fortune 500 companies like IBM (a partner on the ORNL supercomputer construction project along with supercomputer builder Cray.) ORNL’s supercomputer center will bring in additional private market revenue annually from companies co-locating operations for research purposes near such a unique facility.
From attracting more faculty to U.T.-Knoxville to creating an entrepreneurial spirit that will result in spinning off new companies, Zacharia believes ORNL’s unparalleled computational capabilities will, similar to SNS, deliver multiplier economic effects on the whole of Tennessee.
“Increasingly we are relying on an innovation economy to continue to lead the world,” he says. “The next big breakthroughs will be enabled by computing at this scale. That it is located in East Tennessee gives us a huge competitive advantage.”
The future certainly appears bright. Research on tap at ORNL will in all likelihood fulfill on its promise of increased economic activity in Tennessee. All Tennesseans have to do to believe it is to take a moment to study the Pringle, bent in two directions, perfectly placed one upon another, without a single one chipped or broken in the cardboard tube. It may not save the world. But who could eat just one?













