U.S. Supercomputing Needs More Money

STOCKHOLM--The Nobel Foundation plans to sell stock in a new firm being formed to preserve the value of the annual prizes it awards.. . . This year's prizes.. . will each be worth $340,000... . Shrewd investments in the past decade. . . have reversed years of declining value for the prizes, and have raised the foundation's assets to near the real value of the original estate in 1900. --From THE SCIENTIST October 5, 1987, p.4. Lend an ear to hear the story of my galloping success. I was once a Wall Street broker(inside, trader, more or less) But my future brightened quickly Fast enough to lead to shock When the staid Nobel Foundation Said that it would issue stock I began to buy it slowly Craftily, with guile and care Till it soon became apparent I controlled the lion's share Owning fifty-one percent Makes one the boss-as we all know So I called a public meeting And named myself the CEO. The CEO is there to manage Any problems that arise And the main one I adressed was- Who should get the Nobel Prize One who gets it must be brilliant Blessed with competence and In a flash it came upon me If that is what it takes-I fit! First I took the one in Physics But my hunger did not cease So I took the Economics, Medicine then Chem, then Peace, Literature, too, I garnered But thats not where this story ends Since the prize is still awarded I can negotiate with friends. Kritchevsky associate director of the Wistar Institute, 36th St. at Spruce, Philadelphia, PA 19104. The Panel urges the creation of a comprehensive 5-year National Computing Initiative to ensure preeminence of the United States in the development and application of high-performance computing. This requires approximately $1.5 billion in new resources for computational science and engineering over the next 5 years. New opportunities in parallel architectures, microelectronics, optoelectronics, and software, together with the broad enthusiasm of scientists and engineers to apply and integrate high-perfor mance computing technologies, make the National Computing Initiative a technological imperative.... The foremost purpose of this Initiative is to provide significantly improved computational power to U.S. scientists and engineers over the next 5 years and to initiate actions that ensure innovative applications of this powerful technology... The special facilities required to promote the National Computing Initiative will be predicated on the existing National Supercomputer Centers undergoing rapid upgrades of computing equipment and common operating systems . . . In addition the Panel feels that three types of institutions are urgently needed.. . . First, the "computational science and engineering centers" would be devoted to research in computational methods, numerical analysis, algorithm design and development, and related areas such as pattern recognition, imaging, and nonlinear phenomena. Second, the "software institutes" will foster the development of powerful applications and systems in software to the prototype stage, though a mix of basic and applications-oriented research. Third, the "software libraries" will be responsible for software-documentation, maintenance, and dissemination to the computational community. Computational scientists and engineers, computer scientists, programmers, and other researchers interested in developing methods, algorithms, and software to achieve full power of high-performance computing will be members of these centers. All of these institutes and centers would be linked together with the National Supercomputer Centers on the national gigabit network. The Panel unanimously recommends significant increases in support, on a stable basis, for scientific and engineering research that underpins high-performance computing. While architecture and logic design have been and are the driving technology for supercomputers, the Panel urges attention to the increasing importance of developing improved computational methods, and more efficient algorithms to exploit current and future architectures. Progress in these areas will be fostered by the creation of the three types of centers and institutions discussed above, as well as by the existing National Supercomputer Centers. Significant advances must occur in areas such as compilers, programming language, computer design methodology, and operating systems. The Panel urges the development of powerful concepts and technologies for the management of large scientific and engineering databases. The "software institute" and "software library" would contribute to progress in these areas. Several aspects of applied mathematics, such as algorithm design, numerical analysis and the theoretical understanding of nonlinear equations must also advance. The "computational science and engineering centers" will accelerate the solution of these problems. Massively parallel systems are the most promising routes to major improvements in speed and they will require not only more powerful software, but also the availability of reliable, cost effective, and innovative semiconductor device. produced in the United States. Therefore, support for increased basic scientific and engineering studies of advanced semiconductor and other innovative materials, such as high temperature superconductors, is essential to restore the lost US. leadership in the design and manufacturing of semiconductor components. The National Supercomputer Centers, developed by DOE, NASA, and NSF, have provided leadership and visibility and the Panel recognizes the dramatic improvements in access that these centers have been able to achieve in the last 5 years, but is unanimous in noting that access is a continuing problem. Many of the national centers are nearing saturation. Moreover, centers must be rapidly upgraded to have the most advanced computing systems and orders of magnitude improvement in nationwide network facilities. These centers must also reach out to developing fields, such as economics, whose future growth depends on the use of large-scale modeling. Another nationally significant, shortcoming noted by the Panel is the lack of incentive. to have broader segments of US. industry exploit the potential of large-scale computer modeling. It is estimated that thre are approximately 30,000 computer centers nationwide in academe, government, and industry. Representative of the centers that require high- performance-computing should have access to the facilities on the gigabit network. Copies of the report "A National Computing Initiative: The Agenda for Leadership" are available from SIAM, 1400 Arcitects Bldg. 117 S. 17th St. Philadelphia, PA 19103-5053.

U.S. Supercomputing Needs More Money

The Scientist Staff

Interested in reading more?

Become a Member of