UB Today Alumni Magazine Online - Fall 1999
FeaturesAlumni ProfilesClassnotesCalendarThe MailFinal WordEditor's Choice




COVER STORY
Discovering New Media

FEATURES
UB's wired wired world

High-speed Computations

Language Study

Bob Arkeilpane





Related Sites
CCR Website

Pyramid Systems

National Science Foundation

American Crystallorgraphic Association

National Computational Science Alliance

Hauptman-Woodward Medical Research Institute

60 Billion Computations a second


BY ELLEN GOLDBAUM

Innus     When UB faculty and students formed the Graduate Group in Advanced Scientific Computing in 1988, they were simply doing what academics do best: teaming up to explore a field that was then in its infancy.
    Little did they know that they were setting the stage for a research endeavor that would move UB from being virtually nowhere on the supercomputing map to a position of prominence as one of the nation's top 10 academic centers for high-performance computing.
    Together, the supercomputers are able to carry out more than 60 billion operations per second, giving the center the capability of performing a calculation in one day that might take two years to complete on a high-end personal computer.

    Today, visitors to the new Center for Computational Research (CCR) in Norton Hall (www.ccr.buffalo.edu) can see some of the world's most powerful computers at work, quietly carrying out billions of operations per second so that staggering amounts of data can be turned into information scientists can easily use.
    The machines make possible solutions to incredibly complex problems ranging from the design of powerful new drugs and highly efficient aircraft and automotive engines to the development of methods for performing medical diagnoses from remote sites.

    Center users run the gamut from engineers, scientists and physicians to geographers and artists. They also include students and faculty from UB and other SUNY schools, as well as researchers from local companies and research institutions.
    The center features more than $7 million in computers, including IBM's next-generation version of the chess-playing (and winning) Deep Blue computer and the SGI Inc. Origin2000 Server, which is one of the most advanced models of its kind in North America and was used most recently in the creation of the animated movie Antz.
computers     Together, the supercomputers are able to carry out more than 60 billion operations per second, giving the center the capability of performing a calculation in one day that might take two years to complete on a high-end personal computer.
    The center also has state-of-the-art visualization equipment, including the Pyramid Systems' ImmersaDesk, TM which allows users to virtually immerse themselves in an environment, such as a new factory, a waterfront complex, a biological molecule or the inside of an aircraft.

    Although CCR has been in existence just a few months, it is already making major progress:
- The center and Sun Microsystems are partnering to develop a "poor man's supercomputer," based on clusters of workstations that can deliver supercomputing power at a fraction of the cost. This high-visibility, first-of-its-kind project, undertaken by CCR and the university's Science and Engineering Node Services (SENS), casts UB as a member of the product development team for one of the world's major hardware developers. Project leadership at UB is provided by CCR cofounder and director Russ Miller, professor of computer science and engineering, and SENS director Corky Brunskill. The workstations will be connected by a switch provided by CCR partner Nortel Networks.
- Through UB's participation in the National Science Foundation's very-high-speed backbone network service (vBNS), commonly known as the next-generation Internet or Internet2, researchers at the center are involved in high-end projects where data must flow rapidly between remote sites. Active projects include computational steering, in which scientists located at different sites simultaneously participate in and "steer" the design process to a meaningful solution.
- Teams of researchers from all over the world used CCR computers to solve previously unsolvable and biologically important molecular structures during a recent American Crystallographic Association workshop hosted by the center.
- The center is actively pursuing relationships with local industrial partners, particularly those that might want to utilize virtual reality and visualization techniques, whether for designing a new supermarket or making state-of-the-art presentations to clients. IBC, a local firm that develops animation videos for corporate customers, is already using sophisticated tools at the center to complete animations in one weekend that would take them nearly three weeks of computing time to do on their own machines.

    Even in its physical layout, the 4,000-square-foot center is designed to accommodate students and researchers from inside and outside the university, with office space and workstations readily available.
    With these and other initiatives, UB's new center is already positively affecting UB's profile-and not just in high-performance computing.
    According to Larry Smarr, a member of the White House Advisory Committee for High Performance Computing and Communications, Information Technology and the Next Generation Internet, and director of the National Center for Supercomputing Applications (NCSA) at the University of Illinois/Urbana-Champaign, universities are at a critical stage similar to the one they were at after World War II. At that time, major institutions were positioning themselves to take leadership roles in the nation's research enterprise.
    "Those are the ones that still dominate most of the research funding today," says Smarr. "But the Internet revolution is changing all that, and there's a whole new order developing in cyberspace. Universities like UB that understand how to position themselves into this emerging grid could turn over the existing order."
vancomycin     Of course, back in 1988, the Graduate Group in Advanced Scientific Computing wasn't eyeing any such grand plan. Like UB's other graduate groups that formed in the 1970s and 1980s, these faculty and students were simply interested in bringing together researchers from different disciplines on campus to explore opportunities in an emerging field. (UB still has 14 active graduate groups, and new ones are forming all the time.)
    The resources and budget available to the graduate group for advanced scientific computing were modest.
    But because its founders were committed to forging connections across the disciplines and outside of the university, it fostered the kind of research that creates real excitement on a global scale.
    Several major multiyear research projects in computational chemistry were started by members of the graduate group. These projects, now under the direction of Tom Furlani, associate director of the new center; Jiali Gao, professor of chemistry; and Harry F. King, professor of chemistry, are funded by the National Science Foundation, the Environmental Protection Agency and the National Institutes of Health.

    Perhaps the best case in point was the development of a joint research group from UB and the Hauptman-Woodward Medical Research Institute (HWI).
    The effort began in 1988, when the graduate group hosted a seminar with Nobel Laureate Herbert Hauptman, UB research professor of biophysics and president of HWI (formerly the Medical Foundation of Buffalo).
    In his presentation, he described an 18th-century equation called the principle of least squares developed by the German mathematician Karl Friedrich Gauss that was used at the time to predict the orbits of asteroids.
    Hauptman's novel idea was to apply that well-known principle to one of the most persistent problems in crystallography. Its solution, he said, would exponentially accelerate the development of safe, effective new drugs through rational drug design, in which new drug development is based not on trial and error but on detailed knowledge of molecular structures.
    By the time Hauptman presented his idea to the UB Graduate Group for Advanced Scientific Computing, he had already shown that it worked, theoretically; now he needed the tools to prove it could be useful.
    That sparked Russ Miller's interest.
    "At the end of Herb's talk, he put an equation up on the blackboard," Miller recalls. "He said that if we could minimize this equation, then we could solve the phase problem. Since my background isn't in chemistry or crystallography, I knew nothing about the phase problem or its significance, but I knew it was a real-world problem that I wanted to attack."
    It was the start of what would become an extraordinarily fruitful and significant collaboration, one which clearly exemplifies how high-performance computing is changing the way scientists work.
vancomycin     "Russ was able to provide us with ready access to supercomputers, especially through his connections with supercomputing facilities and manufacturers like Thinking Machines," says Hauptman. "The development of algorithms is very computer-intensive. To develop the optimal approach, you have to explore a lot of different directions at once. We have pretty good computers at the institute, but they were by no means adequate for what we had to do."
    Based on Hauptman's minimal function, Miller, along with HWI crystallographer George DeTitta, HWI computational crystallographer Charles Weeks and others devised their "Shake-and-Bake" method for crystal-structure determination, forming the basis of the SnB computer software program.
    With papers published in such top journals as Science and presented at major international meetings, SnB has become the crystal-structure determination software package-of-choice; it has determined large molecular structures in hours or minutes that other methods could not solve even after weeks or months of computation.
    In 1997, after the appearance of a new bacterial strain resistant to vancomycin, "the antibiotic of last resort," a University of Pennsylvania group used SnB to unravel its structure, giving researchers a major weapon against resistant bacteria.
    For computer scientists like Miller, this is the reward of doing computational science: the application of high-performance computing to the development of technical solutions that ultimately benefit society.
    In 1992, the SnB group was rewarded with a four-year, $3 million National Institutes of Health program-project grant; in 1996, it was renewed with an additional $5.6 million. This has been used to continually improve the software to enable it to solve larger and more complex structures.
    "The center will certainly help us with SnB," says Hauptman. "It means that we can seriously think of solving structures with as many as 2,000 atoms-or more. Structures that big could tie up our computing network for a month. A supercomputer can do it in one day."


Ellen Goldbaum is senior science editor with University News Services.

ArchivesGuestbook/FeedbackHomeAlumni HomeUB Home