Calendar No. 710 101st Congress SENATE Report 2nd Session 101-387 HIGH-PERFORMANCE COMPUTING ACT OF 1990 ___________ Mr. HOLLINGS, from the Committee on Commerce, Science, and Transportation, submitted the following R E P O R T OF THE SENATE COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION ON S. 1067 JULY 23 (legislative day, JULY 10), 1990.--Ordered to be printed __________________ U.S. GOVERNMENT PRINTING OFFICE 39-010 WASHINGTON: 1990 COMMITTEE ON COMMERCE, SCIENCE, AND TRANSPORTATION ERNEST F. HOLLINGS, South Carolina, Chairman DANIEL K. INOUYE, Hawaii JOHN C. DANFORTH, Missouri WENDELL H. FORD, Kentucky BOB PACKWOOD, Oregon J. JAMES EXON, Nebraska LARRY PRESSLER, South Dakota ALBERT GORE, JR., Tennessee TED STEVENS, Alaska JOHN D. ROCKEFELLER IV, West Virginia ROBERT W. KASTEN, JR., Wisconsin LLOYD BENTSEN, Texas JOHN McCAlN, Arizona JOHN F. KERRY, Massachusetts CONRAD BURNS, Montana JOHN B. BREAUX, Louisiana SLADE GORTON, Washington RICHARD H. BRYAN, Nevada TRENT LOTT, Mississippi CHARLES S. ROBB, Virginia KEVIN G. CURTIN, Chief Counsel and Staff Director WALTER B. McCORMICK, Jr., Minority Chief Counsel and Staff Director (II) Calendar No. 710 101st Congress SENATE Report 2nd Session 101-387 HIGH-PERFORMANCE COMPUTING ACT OF 1990 JULY 23 (legislative day, JULY 10), 1990.--Ordered to be printed Mr. HOLLINGS, from the Committee on Commerce, Science, and Transportation, submitted the following R E P O R T [To accompany S. 1067] The Committee on Commerce, Science, and Transportation, to which was referred the bill (S. 1067) to provide for a coordinated Federal research program to ensure continued United States leadership in high-performance computing, having considered the same, reports favorably thereon with an amendment in the nature of a substitute and recommends that the bill as amended do pass. PURPOSE OF THE BILL The primary objective of the legislation is to accelerate research, development, and application of high-performance computing in research, education, and industry. High-performance computing is the most advanced computing technology--the most sophisticated computer chips, the fastest computers with the largest memories, the fastest algorithms, and the fastest computer networks. This bill authorizes Federal funding for the development and use of new supercomputers, advanced software, and a National Research and Education Network (NREN), a computer network capable of transmitting billions of bits (gigabits) of data per second. In total, the bill authorizes $650 million for the National Science Foundation (NSF) and $338 million for the National Aeronautics and Space Administration (NASA) for fiscal years (FY) 1991-95. The bill also establishes a National High-Performance Computing (HPC) Program involving NSF, NASA, the Department of Energy (DOE), and the Defense Advanced Research Projects Agency (DARPA) of the Department of Defense (DOD), as well as [2] other agencies. This program would be planned and coordinated by the White House Office of Science and Technology Policy (OSTP). BACKGROUND AND NEEDS IMPORTANCE OF COMPUTING In the last 30 years, computer technology has transformed American science and industry. Today, computers are indispensable tools found in almost every laboratory, office, and factory. They have enabled researchers to solve previously unsolvable problems; have transformed the way products are designed, manufactured, and marketed; have changed the way offices are operated; and have given teachers a new, powerful educational tool. The last five years have seen a rapid increase in the use of supercomputers in science and engineering. Supercomputers are commonly defined as the most powerful computers available at any given time. They usually cost $1-$20 million and are 1,000 to 100,000 times more powerful than a typical personal computer. Today's supercomputers are capable of making billions of mathematical calculations per second, which is about 50 to 100 times faster than the fastest computers available just ten years ago. Using complex computer "models," researchers now can simulate and test the behavior of advanced aircraft designs, proposed new drugs, and new manufacturing techniques. Scientists have used supercomputer models to understand better the Earth's climate and weather, the Nation's economy, the evolution of our galaxy, and even the voting patterns of Members of Congress. To facilitate communication among researchers, students, and educators, and to promote the use of advanced computers, NSF and other Federal agencies have established fiber optic computer networks, which link researchers around the country to supercomputers, to other computing facilities, and to each other. Unlike copper telephone wires, fiber optic cable is capable of carrying the billions of bits of data generated every second by supercomputers. Such high data rates are needed because, for many types of computer models, scientists need sophisticated "visualization" techniques to sort out their results. Computer graphics allow researchers to decipher data sets so large that they could fill hundreds of pages of computer printouts. Unfortunately, most computer networks operate at speeds of 1.5 million bits (megabits) per second or less, and thus network users cannot utilize supercomputers fully. Faster networks also will allow researchers to retrieve huge volumes of data (e.g., satellite images) from data bases and to share their own data with others. Multi-gigabit networks would allow scientists and engineers to control and collect data from research facilities (e.g., particle accelerators and radio telescopes) from thousands of miles away, reducing the need for expensive, time-consuming travel. These high-speed networks would allow researchers around the country to collaborate over the network as effectively as they could face-to-face, leading to the creation of what has been termed a National Collaboratory. In recent years, support has been growing for a large increase in Federal funding for high-performance computing. A November 1985, White House Science Council report, "Research in Very High [3] Performance Computing," states: "The bottom line is that any country which seeks to control its future must effectively exploit high-performance computing. A country which aspires to military leadership must dominate, if not control, high-performance computing. A country seeking economic strength in the information age must lead in the development and application of high-performance computing in industry and research." At a July 21, 1989, committee hearing on his nomination to his current position, Dr. Allan Bromley, the President's Science Advisor and Director of OSTP, stated that high-performance computing must be "a very high priority" because "it has a catalytic effect on just about any brand of research and development" and "will, eventually, transform industry, education, and virtually every sector of our economy, bringing higher productivity and enhanced competitiveness." In a similar vein, a 1989 OSTP report, written by representatives from over a dozen Federal agencies, calls for new funding to "maintain and extend U.S. leadership in high-performance computing." CONGRESSIONAL ACTION Similar interest has been shown in both houses of Congress. In order to spur development of faster computer networks and more advanced supercomputers, in 1986, the Committee reported legislation authorizing NSF programs, which included legislation introduced by Senator Gore, to require OSTP to provide Congress with an analysis of the computer networking needs of American researchers and the benefits and opportunities that a national high-speed fiber optic network for computers and supercomputers would provide. That legislation was enacted into law as part of the NSF Authorization Act for FY 1987 (P.L. 99-383,100 Stat. 816). As required by the legislation, OSTP released a report in December 1987 entitled "A Research and Development Strategy for High Performance Computing," which outlined an ambitious, comprehensive research program in supercomputing and computer networking, and proposed that the Federal Government spend an additional $1.74 billion over the next five years on high-performance computing. This report was followed in September 1989 by an implementation plan for the program, "The Federal High Performance Computing Program," which was developed by more than a dozen agencies working with OSTP. While that report presented a five-year funding profile for a high-performance computing program, the President has yet to endorse the additional funding needed to implement it. However, there are reports that the Administration is preparing a major initiative in this area for FY 1992. The high-performance computing program outlined in the OSTP reports has four elements: high-performance computers; software technology and algorithms; networking; and basic research and human resources. In 1988, S. 2918 was introduced by Senator Gore to create a National HPC Program, similar to that outlined in the 1989 OSTP report. The following year, Senator Gore introduced S. 1067 authorizing funds for high-performance computing at NSF, NASA, DOE, and DARPA. As introduced and as reported this bill differs from the OSTP reports in several ways: it places more emphasis on providing access to scientific data, it seeks to increase industry involvement in a Federal HPC program; it emphasizes more the role of high-performance computing in education; and it specifies funding levels for the different agencies in the program. The first title of the reported bill provides for coordination between the Federal agencies involved in high-performance computing through the Federal Coordinating Council for Science, Engineering, and Technology (FCCSET), which is chaired by the Director of OSTP. In recent years, FCCSET has provided critically-needed, high level interagency coordination of research in a number of areas, most notably global change. The second title of the bill mandates creation, by 1996, of the NREN, a national fiber optic network capable of transmitting billions of bits of data per second from coast to coast. The third title gives NSF responsibility for promoting development of data bases and other information services, which would be available over the NREN. The fourth title provides for development of improved software for supercomputers and other computers. The fifth title funds research and development on new, more advanced supercomputers and related systems. The sixth title calls for more basic research in computing and expanded efforts to educate and train computer scientists and computational scientists (users of high-performance computing). The seventh and final title provides authorizations for NSF and NASA for their contributions to the National HPC Program. To fully reap the benefits of high-performance computing, the Federal Government needs to implement a comprehensive research and development program similar to that provided for in S. 1067. Because the components of the program are all closely linked, progress in high-performance computing will be hindered if the pace of development in any one area is not as fast as in other areas. For instance, if a national high-speed computer network were established, but if faster, more powerful supercomputers were not developed to handle the data that would flow across such a network, the result would be missed opportunities and wasted resources. Similarly, the development of faster supercomputers, without the development of software needed to utilize them effectively and of networks to access them, would be a poor investment of research funds. Clearly, there is a need for a balanced, comprehensive approach. BENEFITS OF S. 1067 Most of the funding authorized in S. 1067 is in support of basic research. There is broad agreement on the general need for the Federal funding of basic research-basic research has been shown repeatedly to be a good investment. For example, in a soon-to-be published study, Dr. Edwin Mansfield of the University of Pennsylvania estimated that the annual rate of return on Federal investments in academic research is approximately 28 percent. The return on investments in basic research on high-performance computing may be even higher. On July 26, 1989, in testimony before the Science, Technology, and Space Subcommittee, Dr. James H. Clark, Chairman and Founder of Silicon Graphics Computer Systems, told how a single $12 million DARPA research [5] grant which Dr. Clark and his colleagues received while he was a professor at Stanford from 1979 to 1982 led directly to the creation of SUN Microsystems, Silicon Graphics Computer Systems, and MIPS Computer Systems. Today, just eight years later, these three computer companies have combined total revenues of almost $2.5 billion per year and an average annual growth rate of 60 percent. In addition, because high-performance computing represents an enabling technology which can increase greatly the productivity not only of computer scientists, but also of researchers in almost all fields of science and engineering, the returns are likely to be greater than the average return on investments in basic research. This research will lead to faster, more powerful computers than can tackle previously unsolvable problems; faster networks that can provide easier access to data and promote collaboration between researchers; and better software that can reduce the time spent computing the solution to a particular problem and thus allow researchers time to explore more facets of a problem. The investment proposed by S. 1067 would provide needed tools for federally-funded researchers and enhance greatly their productivity. At a June 21, 1989, hearing of the Science, Technology, and Space Subcommittee, Dr. William Wulf, then Assistant Director of NSF's Directorate for Computer and Information Science and Engineering, testified that supercomputing and high-speed networking can increase the productivity of many American researchers by 100 percent, 200 percent, or more. Given that the Federal Government invests approximately $70 billion a year in research and development, such a productivity gain could produce enormous benefits and more than pay for the approximately $2 billion total cost of funding the National HPC Program for the next five years. High-performance computing will allow researchers to tackle previously unsolvable problems, with huge benefits to society. For instance, better models of global climate change would lead to better policies to address global warming, policies which could have trillion-dollar impacts. Supercomputing could lead to a better understanding of AIDS, cancer, and genetic diseases, leading to breakthroughs impossible without more computing power. Just as important as the benefits to American researchers are the benefits for American industry. Supercomputers are routinely used by automobile companies, both to design and to "crash test" cars; energy companies use them to analyze seismic data and prospect for oil; and even financial markets now utilize them to get real-time analyses of market behavior. On June 21, 1989, Mr. John Rollwagen, Chief Executive Officer of Cray Research Inc., testified before the Science, Technology, and Space Subcommittee that ARCO used a Cray supercomputer to determine how to increase production of its Prudhoe Bay oil field by two percent, which translates into an additional $2 billion in profits. The engines on Boeing's new 737 airplane were designed using a supercomputer and, as a result, are 30 percent more efficient than earlier models. ALCOA used supercomputer models to reduce the amount of aluminum needed to produce a soda can by 10 percent, resulting in millions of dollars in reduced materials, production, and transportation costs. [6] In the United States, the most extensive use of supercomputers has been for defense and aerospace applications. The National Security Agency (NSA) relies heavily on the fastest supercomputers for signal processing and breaking codes. Supercomputers are essential for anti-submarine warfare and for the design of new weapons systems. The Strategic Defense Initiative and other military research and development projects rely heavily on supercomputer modeling. NASA has several supercomputers devoted to modeling the aerodynamics of aircraft and spacecraft. These supercomputers can be used to replace or complement expensive wind tunnel tests. In the future, high-performance computing will be utilized increasingly by the education and library communities. Supercomputers can store and sort through huge quantities of data, and with optical disk storage systems it is possible to store entire libraries of information electronically and retrieve them in seconds. The Library of Congress and other libraries are starting to develop the technology needed for "digital libraries" of books, journals, images, music, and videos--all stored in digital form and accessible over computer networks. Title III of S. 1067 as reported would provide for expanded efforts to develop such digital libraries. In addition, title III provides for existing Federal data sets, like weather satellite data and census data, to be available on the NREN. At present, a great deal of scientific and economic data is stored in electronic form, but much of it, especially remote-sensing satellite data, is almost inaccessible to researchers and other users. The bill would make data sets like those at the Earth Resources Observation System (EROS) Data Center accessible over the NREN and other networks, thus greatly enhancing the usefulness of these data sets and ensuring that the United States maximizes the return on its investment in the collection of that data. One of the most far-reaching impacts of the bill would be in the area of high-speed, fiber-optic telecommunications technology. Fiber-optic cable can transmit billions and even trillions of bits of data per second, thousands of times more than long-distance copper telephone cables. Scientists and engineers are using this new capability to develop technology for teleconferencing, for using supercomputers and other research equipment remotely, and for improving communication and collaboration among computer users. By creating a national, high-speed computer network, this bill would provide a demonstration of the potential of high-speed fiber optic computer networks. Under this bill, the Federal Government would fund creation of a national multi-gigabit network and development of applications that use it. The technology and standards developed will be available publicly and will be applied quickly by private companies building commercial multi-gigabit networks. At present, the private sector is reluctant to make the multi-billion-dollar investments needed to build a national multi-gigabit network, in part because the technology has not been demonstrated and the market has not been proven. The Federal funding called for in this bill will demonstrate the benefits of a high-speed national network and lead to development of standards for such a network, this stimulating private-sector investment in multi-gigabit networking. At an October 4, 1989, hearing of the House Energy and Commerce Subcom- [7] mittee on Telecommunications and Finance, John Edwards from Northern Telecom testified that Federal funding authorized by this bill could accelerate the creation of a national, multi-gigabit network by 5 to 10 years. Like the interstate freeway system and other types of infrastructure, such a network would provide untold benefits to all sectors of the American economy. HPC AND U.S. COMPETITIVENESS The development of HPC will have a significant impact on U.S. technology competitiveness, particularly given the efforts of other countries to develop a supercomputing capability. The Japanese and other foreign competitors have been quick to recognize the benefits of supercomputing and fiber optic networks. In fact, the Japanese have targeted the world supercomputer market and are now producing some of the fastest supercomputers available. In April 1990, Japan announced a major research program to accelerate research and development on parallel processing supercomputers. Similarly, other countries are making massive investments in high-speed fiber optic networks. Japan's Nippon Telegraph and Telephone Corporation has announced that it intends to invest $126 billion to install a national fiber optic network which would reach every home, office, and factory in Japan by the year 2015 and be capable of transmitting hundreds of millions of bits of data per second. The Europeans are developing initiatives to build their own high-speed networks as part of EC 92. Without additional Federal and private-sector investment in supercomputing, the United States risks losing the $2.4 billion world supercomputer market, and more importantly, it risks having to rely upon foreign suppliers for an essential tool in improving research and development, in increasing American competitiveness, and in enhancing U.S. national security. The funding authorized by S. 1067 would help the United States maintain its lead in the development and application of supercomputers. NEED FOR ADDITIONAL HPC FUNDING To provide supercomputing services to American researchers, the NSF created five supercomputer centers in the mid-1980s. For FY 1991, NSF is requesting $59.59 million to fund the centers. Other Federal agencies, including NASA and DOE, also maintain large supercomputers for use by Federal and academic scientists. The science agencies also fund several computer networks, including NSF's NSFNET, NASA's NASNET and SPAN (Space Physics Analysis Network), DOE's MFENET and HEPNET, and DOD's MILNET. Together with many State-funded or for-profit regional networks, several of these networks are linked by the Internet, which consists of over 2,000 interconnected networks. While it is not known exactly how many computers communicate via Internet, most estimates are that well over 100,000 computers are linked in this way. However, present supercomputing and networking programs are not adequate to meet the needs of researchers. The supercomputers at the NSF centers are chronically over-subscribed. DARPA and other agencies which fund development of new supercomputers [8] lack the money to fund more than a small fraction of the promising proposals for new types of machines, which can cost from $10 million to $500 million to prototype. Furthermore, researchers are often frustrated by the lack of useful research software for supercomputers which stems from the lack of adequate funding for supercomputer software development. Perhaps even more importantly, inadequate funding levels would result in a delay in the establishment of the NREN. The NREN would be capable of transmitting gigabits (billions of bits) of data per second, and by 1996 would link up to 1,300 institutions and about a million researchers nationwide. The NREN would be about 2,000 times faster than the current NSFNET. While this nationwide computer network links over 500 institutions in all 50 States its data rate is only 1.5 million bits per second, more than a thousand times slower than the proposed NREN. Even after NSFNET is upgraded to 45 million bits per second this year, researchers will be unable to utilize fully the supercomputers and data bases connected to it. Since use of NSFNET is growing at a rate of 20 to 30 percent each month, its new capacity will not be enough to accommodate the increased usage expected in the next two or three years. For FY 1991, NSF has requested $22.04 million for NSFNET. Additional funding for both NSF and DARPA will be needed to develop a multi-gigabit NREN. The multi-gigabit NREN is needed if researchers are to use the new networking technology being developed in laboratories around the country. The first computer networks, built in the late 1960s, enabled computers to exchange data at rates of a few thousand bits of data a second (a single page of double-spaced text represents about 10,000 bits of data). Today, there are experimental computer networks that can transmit billions of bits of data a second, enabling computer users to share computer graphics and huge volumes of data in a few seconds. At a billion bits a second, the entire Encyclopedia Brittanica could be transmitted to any computer on the network in less than a second. Unfortunately, these experimental networks are limited, connecting only a few computers. More research and development will be needed before the NREN, which will connect thousands of computers, can be built. According to the President's FY 1991 budget request (p. 85, "Budget of the United States Government)," in FY 1990, the Federal Government spent $448 million on high-performance computing, and for FY 1991, the President is requesting $469 million, an increase of five percent, which barely covers inflation. Without additional funding, researchers will not have access to the supercomputing resources they need; the NREN will be delayed; development of new, more powerful machines will be delayed; supercomputer software development will slow; and insufficient numbers of scientists and engineers will be trained to use supercomputers. The funding authorized by S. 1067 would roughly double funding for supercomputing at NSF and NASA over the next five years, and would roughly triple NSF's networking budget. These funding increases parallel those outlined in the 1987 OSTP report and the follow-up report released in September 1989, "The Federal High Performance Computing Program." [9] LEGISLATIVE HISTORY On May 18, 1989, Senator Gore introduced S. 1067, legislation similar to S. 2918, which Senator Gore had introduced in October 1988. S. 1067 is cosponsored by Senators Jeffords, Durenberger, Pell, Kasten, Conrad, Pressler, Lott, Wirth, Sasser, Kohl, Bryan, Graham, Kerrey, Robb, Gorton, Reid, Kerry, Cranston, Boschwitz, Bingaman, Breaux, and Heinz. The Subcommittee on Science, Technology, and Space held three hearings on S. 1067. The first hearing, on June 21, 1989, focused on the NREN. The second hearing, on July 26,1989, considered the development and application of advanced software and visualization, and the third one, held on September 15, 1989, examined present and future technology for managing and distributing electronic data. Witnesses at the hearings included several leading computer scientists, representatives of computer, supercomputer, and telecommunications companies, and representatives of NSF, DARPA, the U.S. Geological Survey, and the National Library of Medicine (NLM). Witnesses testified on the many new applications of supercomputing, and enthusiastically endorsed the idea of a NREN. Dr. William Wulf, testifying for NSF, stated that the Administration endorses the goals of the legislation, but that it opposes S. 1067 because the funding authorized by the legislation had not been requested in the President's budget. In addition, the Administration believes that a five-year budget for research funding should not be made, because this field is evolving so rapidly. At its April 3, 1990, executive session, the Commerce Committee considered in open session and adopted without objection an amendment in the nature of a substitute for S. 1067. As introduced, S. 1067 provided funding for DOE and DOD, as well as NASA and NSF. Because DOE and DOD programs are not generally authorized by the Commerce Committee, the substitute provides funding for only NASA and NSF. Authorizations for the DOE high-performance computing program are contained in S. 1976, the DOE High-Performance Computing Act, introduced on November 21, 1989, by Senator Johnston and cosponsored by Senators Gore and McClure. The Senate Energy Committee held a hearing on S. 1976 on March 6, 1990, and ordered the bill reported on June 27, 1990. When it was introduced, the sponsors of S. 1976 indicated their intention to have S. 1067 and S. 1976 considered together by the full Senate once each was reported. Funding for the portion of the HPC program to be conducted by DOD's DARPA is expected to be provided in the DOD FY 1991 authorization bill. SUMMARY OF MAJOR PROVISIONS As reported, S. 1067 would authorize a five-year program roughly doubling Federal funding for research and development on supercomputers, advanced computer software, and computer networks. The major provisions are as follows. 1. Title I establishes an interagency National HPC Program involving NSF, NASA, DOE, DOD, and other relevant agencies. Interagency coordination and planning for the program would be [10] provided by OSTP's FCCSET, which is to work closely with industry. 2. Title II requires NSF to work with other agencies to establish a multi-gigabit NREN by 1996. This network would be capable of transmitting several billions of bits of data per second and would link hundreds of thousands of researchers in government, industry, and universities around the country. 3. Title III makes NSF the lead agency for ensuring that federally-funded data bases and network services can be accessed over the network. 4. Title IV calls for expanded software research and development, especially on software for supercomputers. Most of the funding would go to scientists, engineers, and computer scientists using high-performance computing to solve so-called "Grand Challenges," fundamental problems in science and engineering, examples of which are provided in the bill as reported. 5. Title V provides for research on supercomputers and encourages development of new supercomputing technology by the private sector. 6. Title VI requires OSTP to oversee and coordinate Federal programs for basic research in computer technology and for the education of computer scientists, computational scientists, information scientists, and electrical engineers. 7. Title VII authorizes $338 million to NASA for FY 1991-95 for the purposes of the bill. For FY 1991-95, the bill authorizes for NSF $195 million to establish the network, $64 million for basic research and education, and $391 million for the other purposes of titles III, IV, and V. The total authorization for NASA and NSF for FY 1991 is $68 million. ESTIMATED COSTS In accordance with paragraph II(a) of rule XXVI of the Standing Rules of the Senate and section 403 of the Congressional Budget Act of 1974, the Committee provides the following cost estimate, prepared by the Congressional Budget Office: U.S. CONGRESS, CONGRESSIONAL BUDGET OFFICE, Washington, DC, April 27, 1990. Hon. ERNEST F. HOLLINGS, Chairman, Committee on Commerce, Science, and Transportation, U.S. Senate, Washington, DC. DEAR MR. CHAIRMAN: The Congressional Budget Office has prepared the attached cost estimate for S. 1067, the High-Performance Computing Act of 1990. If you wish further details on this estimate, we will be pleased to provide them. Sincerely, ROBERT D. REISCHAUER, Director. CONGRESSIONAL BUDGET OFFICE COST ESTIMATE 1. Bill number: S. 1067. 2. Bill title: The High-Performance Computing Act of 1990. [11] 3. Bill status: As ordered reported by the Senate Committee on Commerce, Science, and Transportation, April 3, 1990. 4. Bill purpose: S. 1067 would require the Federal Coordinating Council on Science, Engineering, and Technology (FCCSET) to develop and implement a National High-Performance Computing Plan. It would also mandate that the National Science Foundation (NSF), in conjunction with the Department of Defense (DOD), the National Aeronautics and Space Administration, and other relevant federal agencies, establish a national network of high speed computers, which would be known as the National Research and Education Network (NREN). The requirements of S. 1067 would affect numerous federal agencies as developers and users of the NREN. For example, the bill would requires the NSF and NASA to help develop software for the types of computer used in the network; the National Institute of Standards and Technology, a part of the Department of Commerce, would be charged with developing government-wide standards for computer networks; and the DOD, through the Defense Advanced Research Projects Agency (DARPA), would have primary responsibility for research and development on technology needed for the network. The FCCSET would have the general responsibility for overseeing and coordinating the work of the agencies involved in this project. To fund development of the network, the bill would authorize appropriations to the NSF and NASA of nearly $1 billion over five years. The bill would also authorize NSF to charge a fee for use of the system. 5. Estimated cost to the Federal Government: (By fiscal year, in millions of dollars) 1991 1992 1993 1994 1995 _________________________________________________________________ Specific authorizations: National Science Foundation. . . 46 88 145 172 199 NASA. . . . . . . . . . . . . . . 22 45 67 89 115 Subtotal, specific authorizations 68 133 212 261 314 Estimated authorizations. . . . . 12 13 13 14 14 Total authorization level. . . . .80 146 225 275 328 Estimated Outlays . . . . . . . . 45 107 180 241 295 _________________________________________________________________ The costs of this bill would be in budget functions 250 and 370. Basis of estimate: This estimate assumes that the full amounts authorized would be appropriated for each fiscal year. Based on information provided by the Department of Commerce, CBO estimates that the cost of research and studies required by the bill but not specifically authorized would be roughly $12 million per year beginning in 1991; this amount, with adjustments for inflation, is shown as the estimated authorization level in the table above. The estimated outlays are based on historical spending patterns. CBO expects that fees for use of the network would be phased in once the network is operating, which would probably be in 1993 or 1994. Receipts from these fees could ultimately provide a signifi- [12] cant offset to the operating costs of the network. Nevertheless, we do not expect that receipts would be significant during the five-year period covered by this estimate. 6. Estimated cost to State and local governments: None. 7. Estimate Comparison: None. 8. Previous CBO estimate: None. 9. Estimate Prepared by: Doug Criscitello and Michael Sieverts. 10. Estimate approved by: James L. Blum, Assistant Director for Budget Analysis. REGULATORY IMPACT STATEMENT In accordance with paragraph II(b) of rule XXVI of the Standing Rules of the Senate, the Committee provides the following evaluation of the regulatory impact of the legislation, as reported. NUMBER OF PERSONS COVERED This legislation provides additional funding for research and development in high-performance computing. This will not result in new regulations, because the additional funding provided by the legislation would be distributed according to existing regulations regarding NSF research grants and NASA contracts. These regulations would apply only to those persons and companies choosing to apply for this funding. ECONOMIC IMPACT This legislation authorizes $988 million in additional Federal spending for FY 1991-95. By providing for improved inter-agency coordination, this legislation should improve the effectiveness of Federal research and development on high-performance computing. This legislation also requires the National Institute of Standards and Technology (NIST within the Department of Commerce (DOC) to develop guidelines and standards: (l) to provide for interoperability of Federal computer networks, and (2) to promote the use of "open systems software" which can run on several different computer systems. These guidelines should be cost-effective in increasing the usefulness of Federal networks and software purchased by the Federal Government. PRIVACY This legislation will not have any adverse impact on the personal privacy of individual Americans. The creation of the NREN and associated databases will make existing Federal scientific data bases (including economic data and census data) more accessible to users throughout the country, but personal data already protected by rules and regulations (e.g., tax returns and individual census forms) will remain confidential. PAPERWORK This legislation requires FCCSET to submit an annual report to the President and the Congress on the National HPC Program. The National Telecommunications and Information Administration (NTIA) within DOC is to report to Congress on whether State and [13] Federal telecommunications laws and regulations hinder or facilitate private industry participation in the data transmission field. OSTP shall report to Congress on options for charging users of the NREN and the databases connnected to it. DOC is to report to Congress on whether Federal procurement regulations discourage the development of better software development tools and on whether export controls hinder the development of foreign markets for North American manufacturers of high-performance computer systems. SECTION-BY-SECTION ANALYSIS SECTION 1.--SHORT TITLE This section states that the bill may be cited as the "High-Performance Computing Act of 1990". SECTION 2.--FINDINGS AND PURPOSES This section contains the Congressional findings and purposes of the Act. Under subsection (1), Congress finds, among other things, that advances in computer technology are vital to the prosperity, national and economic security, and scientific advancement of the United States. Under subsection (b), to maintain leadership in computer technology and its applications, and to reap the benefits of high-performance computing, the reported bill calls for expanding Federal support for research, development, and application of high-performance computing, and improving planning and coordination of Federal research and development on high-performance computing. Subsection (b) states that this legislation is intended to help establish a high-capacity national research and education computer network; expand the numbers of researchers, educators and students with training in and access to high-performance computing; develop a system of data bases and other services available through such a network, accelerate development of more powerful supercomputers and other advanced computer systems; stimulate research and development of better software for both supercomputers and other computers; promote application of high-performance computing to "Grand Challenges" of science and engineering; and provide for basis research in high-performance computing. SECTION 3.--DEFINITIONS Definitions of "North American Company" and "Grand Challenge" are provided in Section 3. Several provisions in this bill are designed to assist U.S. industry. However, the recently signed Free Trade Agreement between the United States and Canada is to encourage economic cooperation between the two nations and reduce economic barriers. Therefore, the term "North America company" is used to allow Canadian and U.S.-Canadian joint ventures to benefit from these provisions. A "Grand Challenge" is a fundamental problem in science and engineering, with broad economic and scientific impact, whose solution will require the application of high-performance computing resources. Examples of Grand Challenges include modeling of global [14] change, designing of new materials and drugs, and deciphering of the human genome. TITLE I SECTION 101.--NATIONAL HIGH-PERFORMANCE COMPUTING PROGRAM This section amends the National Science and Technology Policy, Organization, and Priorities Act of 1976 (42 U.S.C. 6601 et seq., hereinafter referred to as "the Science Act") in order to establish a National HPC Program coordinated by OSTP. This section would add a new title VI to the Science Act, with the following sections: New section 601 contains findings similar to those in section 2 of the reported bill. New section 602 mandates a National HPC Plan, which under subsection (a)(l) is to be developed and implemented by FCCSET. FCCSET is chaired by the Chairman of OSTP, who is traditionally also the President's Science Advisor. It is currently charged with addressing research issues and coordinating research programs that involve more than one Federal agency. For instance, in recent years, FCCSET has done an exemplary job of providing high-level coordination on global change research. The National Global Change Research Program can serve as a model for the National HPC Program. As with global change research, high-performance computing involves several agencies, and there is no one agency with the expertise, breadth, and facilities to oversee all Federal efforts in the field. FCCSET provides a mechanism for building on existing agency programs, preventing duplication of effort, and identifying previously unaddressed problems, without establishing a new bureaucratic entity. In addition, building on existing agency programs, rather than creating a separate agency for high-performance computing, would ensure that new developments in high-performance computing are utilized by individual agencies to accomplish their different missions. Under subsection (a)(2) of new section 602, the plan is to establish the goals and priorities for a National HPC Program, to set forth the roles and computer research budgets of the agencies involved, and to include the results of studies by Federal agencies and departments, the National Research Council (NRC), and other entities. The Committee expects that the plan would be similar to the 1989 OSTP report, "The Federal High-Performance Computing Program," except that it also would include a budget showing the level of funding for each of the activities undertaken in support of the program by each of the agencies involved and provide a comprehensive inventory of what high-performance computing programs are currently underway throughout the Federal government that could contribute to the National HPC Program. Under subsection (a)(3), the plan would summarize the activities of NSF, DOC, NASA, DOD, DOE, the Department of Health and Human Services (HHS), the Department of Education, the Library of Congress, the NLM, the National Agricultural Library, and other relevant agencies. Subsection (b) of new section 602 provides that FCCSET will have the lead in developing and implementing the plan. At least once a [15] year, the Chairman of FCCSET will report to the President on how to improve implementation of the plan. Working through the Executive Office of the President, and especially with the Office of Management and Budget (OMB), FCCSET should provide the high-level coordination needed to direct and implement effectively the National HPC program. Coordination between OMB and FCCSET will be critical to the success of the program. Under subsection (b)(3), prior to the submission of the President's annual budget request, FCCSET will review each agency and department's budget estimate to determine how it contributes to the implementation of the HPC program. This review is intended to guide OMB in determining each agency's budget for high-performance computing. In addition, FCCSET will work with Federal agencies, the NRC, and other groups involved in high-performance computing to formulate and implement the plan. To receive advice from industry and academia, OSTP is directed under subsection (c) of new section 602 to establish an HPC Advisory Panel which will consider issues pertaining to the program and evaluate its progress, focus, and direction. This panel primarily should provide advice on policy decisions regarding the National HPC Program. The Committee realizes that additional funding will be necessary for this panel and encourages the Administration and the Appropriations Committees to increase funding for OSTP accordingly. This new section in subsection (d) also describes the existing missions of ten of the agencies that will contribute to the National HPC Program. By summarizing the agencies' existing roles, the Committee intends to ensure that existing agency programs are incorporated into the National HPC Program. The inclusion of current research roles is not intended to limit agency responsibilities or to interfere otherwise with the flexibility that will be required by FCCSET to develop and implement a comprehensive national research program. The Committee expects agency responsibilities to evolve over time, as the challenges associated with high-performance computing change and the technology advances. In addition, this subsection requires the agencies to work together to connect their computer networks and to improve, share, and distribute computer software they have developed. As outlined in the 1989 OSTP report, the four key agencies in the National NPC Program are NSF, DARPA, NASA, and DOE. NSF will continue to fund university research and development in high-performance computing and provide researchers with access to state-of-the-art supercomputers. In addition, NSF is to implement the NREN. DARPA's first priorities will be to develop the multi-gigabit network technology needed to build the NREN and to fund development of future generations of supercomputers and other advanced computer systems. DARPA will fund development of new advanced conventional and massively-parallel supercomputers. DARPA also will work with and provide technical support to the high-performance computing research programs within the Army, the Navy, the Air Force, the National Security Agency, and other parts of DOD. NASA will continue to conduct basic and applied research on high-performance computing, with particular emphasis on the development of applications for supercomputing in [16] areas like aeronautics and the processing of remote-sensing data. DOE will continue to conduct basic and applied research in high- performance computing, particularly in software development and development and use of parallel-processing supercomputers. Concerns have been expressed in the past that technology developed at DOE has not been transferred effectively to the private sector. As part of the National HPC Program, DOE and other agencies will ensure that unclassified computer research is readily available to North American companies. This provision should not be interpreted as barring foreign firms from access to high-perfo