Study eyes terascale road map

Millimeter-sized sensors could one day give every frog an IP address. From those frogs, mass amounts of data could be collected and analyzed with ecological models to influence environmental public policy, according to the director of one of the nation's top supercomputing facilities.

The idea of giving every frog a home on the Internet, as described by Dan Reed, director of the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, may seem far-fetched. But the vision for a highly integrated network of instruments, sensors, supercomputers and high-speed communications that could create opportunities such as the frog scenario is the subject of a National Science Foundation study due out this summer.

NSF established the Advisory Committee for Cyberinfrastructure in late February to design a road map for the agency to connect terascale resourcesat facilities and universities across the country. The committee will comprise about a dozen members from academia, industry and government.

The committee is expected to report how the research instruments, supercomputers and high-performance networks the agency supports could be integrated into a more powerful cyberinfrastructure.

That cyberinfrastructure includes not only advanced computing systems and high-performance networks, but also the databases, software and devicesneeded for collection and visualization of vast datasets.

NSF's fiscal 2001 budget for those types of tools across its science directorates is $138.5 million. That includes millimeter-array radio telescopes, a network of earthquake engineering simutors, a South Pole network and a future distributed network of teraflop processors. A teraflop equals 1 trillion floating-point operations per second.

The tools all share these features: They generate terabit data and the need to display it, and they generally are geographically distributed or else their users are, said Ruzena Bajcsy, NSF assistant director of the Directorate for Computer and Information Science and Engineering (CISE)."What doesn't exist today is the integrated system."

Besides bridging the technological divides among the components, NSFalso needs to assess how to bridge the cultural gaps between the various users and developers of high-end computing instruments and applications, Bajcsy said.

NSF intends to use the advisory committee's report to create a five-year program that would serve all the sciences, Bajcsy said. "It will produce a different kind of science than we do now because it is now fragmented."

A similar committee was created in 1995 to help NSF migrate from supercomputing centers to supercomputing partnerships, said George Strawn, executive officerof CISE. The result was NSF's Partnerships for Advanced Computational Infrastructure(PACI), which is now soliciting proposals for a $45 million distributed terascale system — "a first step toward the cyberinfrastructure."

Reed, who is also director of the National Computational Science Alliance, one of the PACI partnerships with NSF, said the study is a natural progression of the PACI program because it moves beyond islands of supercomputing.

The cyberinfrastructure vision takes research a step beyond joining experimental facilities by adding a dimension of collaborative technology to allow distributed groups to work together on large-scale problems, Reed said.

Strawn said he hopes there will be adequate budget resources for the cyber-infrastructure work. NSF received a $56 million increase above its2001 level in President Bush's budget released last month.