GovernmentExecutive.com

On the surface, all would appear to be well with the U.S. supercomputing program. In the past month, two U.S. supercomputers broke the calculation speed record. The National Aeronautics and Space Administration and Silicon Graphics announced in late October that their new supercomputer, "Columbia," named for the space shuttle destroyed on re-entry last year, was the most powerful in the world after it reached a sustained performance of 42.7 trillion calculations per second, or teraflops.

Then, two weeks ago, the Energy Department and IBM told the world that their "Blue Gene/L" supercomputer, developed for nuclear weapons research, had eclipsed that record with a performance of 70.72 teraflops. At the time, the machine was running at only one-quarter of its final capacity. Before October, the record-holder since 2002 was Japan's "Earth Simulator," which was rated at 35.86 teraflops.

Although these dizzying calculation speeds are important, experts in the computer industry say that America needs to do more to keep its technological edge. They call for better access to supercomputers across industries and research fields, so that more than a few institutes or high-end players benefit. Advocates say that the country must also develop a long-term investment plan and a "road map" to maintain its edge in the supercomputing field. Crucial to all of this, they say, is a boost in government leadership and investment.

"Whether the most powerful and most expensive supercomputer is located in the United States or elsewhere does not indicate loss or gain of leadership in supercomputing technology," said Marc Snir, a professor of computer science at the University of Illinois (Urbana-Champaign). "Rather, our concern is that current investments and plans are not sufficient to provide the capabilities that our country needs."

High-performance computers, whether they are enhancing national security, predicting natural disasters, or making a success of Shrek 2, are critical to U.S. global competitiveness.

"Government has traditionally played an important role in developing new technologies," said Suzy Tichenor, a vice president at the Council on Competitiveness, and the director of its high-computing initiative. "I think what we're seeing now is a need to reinvigorate that partnership role." The council is a group of corporate CEOs, university presidents, and labor leaders working for policies to keep the United States competitive internationally.

Over the past year or so, several reports have pointed to the need for a road map for U.S. supercomputing efforts, and for stable, long-term research funding. A National Research Council report released last week criticized government efforts as inadequate. "U.S. leadership is eroding, and with it, the broader scientific and technological progress underlying a strong and robust economy," said Susan Graham, a professor of computer science at the University of California (Berkeley), who co-wrote the report with Snir.

"It is critically important, both for economic security and national security, to retain our leadership position in high-performance computing," said David Shaw, former science adviser to President Clinton and founder of a multibillion-dollar investment and technology development firm, D.E. Shaw. "This will require a substantial sustained investment, but the returns could be far larger." Shaw, a member of theCouncil on Competitiveness executive committee, is working to develop a supercomputer for biomedical applications.

"If the United States is going to be able to compete effectively on a global basis, the economy has to be built on a foundation of innovation," said Tichenor, noting that the rest of world is catching up to the U.S. in the availability of capital, talent, and technology. Tichenor's message: "High-performance computing is an innovation accelerator."

Last week, Pittsburgh hosted SC2004, the High-Performance Computing, Networking, and Storage Conference, which is the biggest yearly event bringing together high-performance computer-makers, vendors, and users. The focus was not only on the latest technologies, but also on strategies for making them more accessible.

David B. Nelson, director of the government's National Coordination Office for Information Technology Research and Development, said in an interview recently that supercomputing "has tremendous promise, but it's up to us to realize that promise."

Government support for supercomputing research began in the 1960s and continued until the 1990s, when government pulled back somewhat. Then in 2002, Japan's announcement that its "Earth Simulator" had broken the U.S.-held world-record computing speed served as a warning for the American science policy community.

The U.S. government remains the world's largest purchaser of supercomputers, mostly for use in national security or climate modeling. But experts say the government should find ways to share its expertise with industry.

The government seems to be getting the message. Several federal agencies have increased their involvement in supercomputing, and the White House Office of Science and Technology Policy, led by Director John Marburger, has begun to take a leadership role. Marburger's office oversaw the formation of a High End Computing Revitalization Task Force, which in May issued a plan and a five-year road map for federal investments beginning in fiscal 2006.

But government funding has not increased significantly in recent years. The fiscal 2005 White House budget calls for approximately $2 billion for the networking and information technology research-and-development program, a modest increase of 14 percent since 2001. Marburger has said that high-performance computing will be a priority in the fiscal 2006 budget, but given federal budget deficits, some analysts see tough times for research spending.

Meanwhile, the Defense Advanced Research Projects Agency, the Pentagon's incubator for society-changing technologies such as the Internet, has renewed its involvement in the field, funding a project aimed at bridging the gap between government and commercial computing capabilities. At the Energy Department, Science Office Director Ray Orbach is called a "visionary" by many in the field.

Congress is also stepping up. This week, lawmakers passed and sent to the president H.R. 4516, sponsored by Rep. Judy Biggert, R-Ill., which requires the Energy Department to establish and operate a leading high-end computing facility. The bill authorizes an additional $50 million in fiscal 2005, $55 million in 2006, and $60 million in 2007.

A more comprehensive bill, H.R. 4218, passed this year by the House, is expected to be reintroduced in the new Congress by House Science Committee Chairman Sherwood Boehlert, R-N.Y. The bill would update the 1991 High Performance Computing Act by strengthening interagency coordination and giving the White House more authority. The bill requires the White House Science Office to create a technology development road map and an investment plan for federal high-end systems.

"Supercomputing has become an essential resource for U.S. industry and academia," Boehlert told National Journal recently. "The House Science Committee will continue to press to ensure that U.S. scientists and engineers have access to the supercomputing capabilities they need to remain world leaders."

Marburger, whose office would be more accountable for supercomputing efforts under the bill, in June charged the President's Information Technology Advisory Committee, or PITAC, with assessing federal research in computational science.

Daniel Reed, chair of the PITAC subcommittee on computational science, gave a progress report at an early-November subcommittee meeting. He said that weather and climate is a key application for supercomputing, because weather can affect 40 percent of the $10 trillion U.S. economy. Currently, he said, natural-disaster predictions often lead emergency authorities to "over-warn" -- by a geographical factor of three, or more than 200 miles -- civilian populations that might be in the path of a hurricane. As a result, evacuations affect many more people than really necessary, costing the economy an average of $200 million extra per occurrence, he said.

Reed identified other problems that his subcommittee has found in its research: a disconnect between the needs of business and the computing infrastructure needs of government and academia; a lack of long-term investment in supercomputers; too little interagency coordination; and inadequate supercomputing software.

To be sure, not everybody is convinced of the wonders of supercomputers. Rep. Brad Sherman, D-Calif., inserted a provision in the House version of the Energy Department's bill that would require research into the societal implications of supercomputing.

Sherman said in an interview that he is concerned that research in several cutting-edge scientific areas, including high-end computing, could lead to the creation of a new race of beings more intelligent than humans. "We're 25 years away from a computer that can justifiably ask for the minimum wage," he said, adding with humor, "As a Democrat, I think that should be a higher minimum wage."

Sherman contended that researchers should ask not only how to develop the technologies, but whether it is desirable to do so. The last time a new race came along, he said, was "when the Cro-Magnon said hello to the Neanderthal -- and that didn't work out so well for the Neanderthal."

Ultimately, the Senate Energy and Natural Resources Committee dropped Sherman's provision, but only because it would have required approval from another committee, he said. Sherman will continue looking for opportunities to bring his measure up, an aide said.

But Sherman will get a big argument from the Council on Competitiveness, which recently reported the results of a survey finding that industry executives consider supercomputing crucial to their companies' survival. Some executives said that more-powerful and easier-to-use supercomputers could save billions of dollars. Using robust computers, researchers and companies can run complex simulations and models more quickly. "High-performance computing really becomes a baseline resource" for all economic sectors, Tichenor said.

Examples of high-end computing abound in defense, intelligence, transportation, health care, pollution control and remediation, and fuel efficiency. A typical desktop computer today has the power of a supercomputer from a decade ago.

Shrek 2's cutting-edge computer-generated animation, for example, was produced using high-performance computing. But India and China are already racing to catch up in computer-generated animation. Top foreign computer-science graduates of American universities are increasingly being drawn back to their home countries. And Europe last week announced the "birth" of a continent-wide supercomputing facility that is globally competitive.

That's why U.S. experts cheered the good news of the computing speed accomplishments of "Blue Gene/L" and "Columbia" -- achievements that make America look like a leader. After Silicon Graphics's record-breaking announcement, CEO Bob Bishop predicted a "trickle-down effect" throughout the economy and declared that top tech jobs would remain in the United States. "This machine will inspire young emerging engineers and scientists across the nation to come back into science, come back into engineering, and pursue this outstandingly exciting era."

Post a Comment

To post a comment, you must provide a name and a valid e-mail address. Messages must be limited to 400 words. By using this Service you agree not to post material that is obscene, harassing, defamatory, or otherwise objectionable. Although Government Executive does not monitor comments posted to this site (and has no obligation to), it reserves the right to delete, edit, or move any material that it deems to be in violation of this rule.