Wednesday, November 27, 2002

'Grid Computing' Is the Next Wave in High-Performance Computing By FLORENCE OLSEN After Internet2, "grids" may be the next big thing in high-performance computing. A grid is a network of computational research centers whose supercomputer clusters, databases, and specialized programs form a pool of resources that is more powerful and more versatile than that of any single research center, or node, on the network. An example would be North Carolina's BioGrid, a statewide infrastructure that university scientists will use for conducting genomic research. Universities that acquired computing clusters in the past few years are moving now to the next stage and starting to create grids, says Rick Herrmann, industry manager for high-performance computing at the Intel Corporation. Grids enable different universities to pool their computing resources for solving large science problems. Understanding black holes in the universe is one such problem. Regional tornado forecasting is another. Grid software may well change the way scientists and, eventually, the rest of us interact with computers, according to some academic researchers who work with grid technologies. Grid experts expect to create a simple Web interface that will make the computational part of science -- searching remote databases or running programs on a distant computer -- as easy as using a utility service such as that provided by electrical power grids. Proponents say grid computing could have as big an impact as the Web has had on who uses the Internet and how they use it. Still, it won't happen overnight. "The full revolution will be many years in the making," says Ian Foster, a professor of computer science at the University of Chicago who is a senior scientist at the Argonne National Laboratory. Mr. Foster and Carl Kesselman, director of the Center for Grid Technologies at the University of Southern California's Information Sciences Institute, are the principal architects of Globus, the basic grid software that has become the de facto standard for grid computing. At least 18 grids for research and education have been built or are under construction, including one very large grid in China. The Chinese Ministry of Education wants to put 100 universities in China on a grid within three years, says Mr. Herrmann, of Intel. Institutions and even some countries are in a race to see which one of them can build the best grid for life-science research, such as genomics and bioinformatics. "The countries that build out the best infrastructure to support their research and science are the countries that are going to draw the best intellectual capital to solve the problems," Mr. Herrmann says. An example of such infrastructure is the Extensible Terascale Facility. In 2003, scientists and engineers will begin using that grid, which network-computing experts say will be the world's largest and fastest grid for open, scientific research. The National Science Foundation is spending $88-million on the research grid, whose computing power will be on the order of 20 trillion operations per second. Three supercomputing centers, a prominent research university, and a national research lab will be mere nodes on the Extensible Terascale Facility. They are the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign, the San Diego Supercomputer Center at the University of California at San Diego, the Pittsburgh Supercomputing Center at Carnegie Mellon University and the University of Pittsburgh, the Center for Advanced Computing Research at the California Institute of Technology, and Argonne National Laboratory, near Chicago. The network for the terascale grid will whip data around at peak speeds of 40 billion bits per second. That is four times faster than the Internet2 backbone will be when its upgrade to 10 billion bits per second is completed in 2003. The grid network, however, will not replace the Internet2 backbone network that now links the five research facilities. North Carolina may be the first state to build its own terascale computing grid for life-science research. The inauguration of the statewide North Carolina BioGrid, scheduled for the middle of next year, is expected to accelerate the pace of drug discovery and agricultural research and to be a boon to the state's economy. The statewide BioGrid, which will be smaller than the science foundation's Extensible Terascale Facility, will have the power of a supercomputer capable of 1 trillion operations per second. In the lineup as nodes on the grid are the North Carolina Supercomputing Center at Research Triangle Park, Duke University, North Carolina State University, and the University of North Carolina at Chapel Hill. About $1.5-million has been allocated so far for the research grid. According to Jack DaSilva, a computational scientist at the North Carolina Supercomputing Center, a grid will be useful for molecular biologists. "If you bring together a half dozen sites," he says, "the range of resources available is going to be greater than any one institution is probably going to make available to users." Mr. DaSilva adds that biologists and molecular biologists "haven't actually said that they need a grid." But, he says, that probably is because "the only people that really know very much about grids right now are the people that are in information technology developing them." Easy-to-print version E-mail this article

Many colleges could close or merge because of financial problems, Standard & Poor's warns OPINION U. of Virginia dismisses 45 students and revokes 3 diplomas as cheating probe concludes Capital campaigns are announced at 5 colleges and universities "Grid computing" is the next wave in high-performance computing