Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 1
INTRODUCTION As requested by the Defense Authorization Act for FY 1994, the National Research Council’s Computer Science and Telecommunications Board organized a broad-based study committee (see appendix) to assess the status of the HPCCI. The act specified submission of an interim report by July 1, 1994, and a final report by February 1, 1995, to the secretary of defense and the director of the Office of Science and Technology Policy. Congressional concerns, which were reflected in the charge to the committee, included the HPCCI’s intellectual directions, rationale, goals, private-sector input, balance, management, and progress, and the relationship of the initiative to other federal support of high-performance computing and communications. The technical experts and congressional staff who have met with the committee to date have focused on the high-performance computing portion of the HPCCI program, an emphasis that this interim report reflects. The committee’s final report will also address the likelihood that the initiative will achieve its goals, several issues of balance, the relationship of the initiative to other federal high-performance computing and communications activities, and the role of the HPCCI within the National Information Infrastructure (NII) initiative. THE HIGH PERFORMANCE COMPUTING AND COMMUNICATIONS INITIATIVE “The goal of the Federal High Performance Computing and Communications Initiative is to accelerate the development of future generations of high-performance computers and networks and the use of these resources in the Federal government and throughout the American economy.”1 Since its formal inception in 1991, the HPCCI has attained considerable visibility both within the computer research community and within the administration’s technology-related plans and programs. When originally formulated, the HPCCI was aimed at meeting several “Grand Challenges” such as the modeling and forecasting of severe weather events. It was subsequently broadened to address “National Challenges” relating to several important sectors of the economy such as manufacturing and health care. The current administration is also relying on the initiative to help develop the technology base needed for the National Information Infrastructure. The HPCCI budget has grown from $489.4 million in FY 1992 to the $1.1 billion requested for FY 1995. This broadening of the federal framework for computing and communications has raised questions about the HPCCI’s goals and may have led to expectations that were not embodied in the initiative as originally proposed.
OCR for page 2
Development and Structure Beginning in the early 1980s, several federal agencies developed independent programs to advance many of the objectives of what was to become the HPCCI. The Office of Science and Technology Policy assumed coordination of these activities, acting through the Federal Coordinating Council for Science, Engineering, and Technology and two of its subgroups, the Committee on Physical, Mathematical, and Engineering Sciences and its High Performance Computing, Communications, and Information Technology (HPCCIT) subcommittee. The program received added impetus and more formal status when Congress passed the High-Performance Computing Act of 1991 (PL 102–194) authorizing a five-year program in high-performance computing and communications. This act affirmed the interagency character of the HPCCI, assigning broad research and development emphases to the 10 federal agencies that were then participating in the program without precluding the future participation of other agencies. Public Law 102– 194 also laid the foundations for some of the eventual broadening of the HPCCI and expanded the concept of the networking component, which became known as the National Research and Education Network program. The HPCCI fosters collaboration by recognizing common goals, reducing administrative barriers, and providing a forum for communication among agencies. The diverse agencies in the HPCCI program retain their individual missions and invest in computing activities relevant to those separate missions. The HPCCI does not usurp agency authority by imposing central control. The organizational structure of the HPCCI has evolved steadily, largely in response to external pressures for improved visibility of decision making, requirements for accountability for expenditures, and the flow of information into and out of the initiative. In September 1992, the National Coordination Office (NCO) for High Performance Computing and Communications was established to aid interagency cooperation and to serve as liaison for the initiative to the U.S. Congress, other levels of government, universities, industry, and the public. The NCO assists the mission agencies in coordinating their separate programs, offering a forum through which the separate agencies can learn of each other’s needs, plans, and actions. As part of its coordinating function, the NCO gathers information about the HPCCI activities of different agencies and helps to make this information available to Congress, industry, and the public. Since its formation, the NCO has published the FY 1994 and FY 1995 “blue books” describing the HPCCI, its accomplishments, and future objectives.2 Technical Background and Progress The technical developments and issues outlined below indicate the larger context in which the HPCCI is best understood. During the 1980s, computing in the United States underwent a revolution and spread rapidly beyond its traditional boundaries: individuals increasingly used ever more powerful personal computers in their homes, schools, and businesses.3 Many segments of government, industry, finance, science, medical research, and engineering regularly used the fastest of the high-performance computers, so called supercomputers, to solve complex problems. Computer networks and other forms of reliable, high-speed digital communications moved—in varying degrees—from the computer research community into the nation’s commercial telecommunications network, and thus into general use.
OCR for page 3
Hardware and Networks Much of this progress was built on rapid advances in integrated circuit technology. By the late 1970s, integrated circuit design and manufacturing had become sophisticated enough that an entire computer could be built on a single chip, leading to the development of the microprocessor. Integrated circuits had also become faster, cheaper, and more reliable than the discrete technologies they replaced. Both industrial and federally funded research on the design and manufacture of very large scale integrated (VLSI) silicon chips had contributed to this progress. The 1980s saw astounding advances in the power and speed of these microprocessors; for most of the decade, the number of transistors per dollar, or the computing power per dollar, doubled approximately every 18 months.4 At the same time, digital networking on a truly global scale emerged from the packet-switching network experiments directed and funded by ARPA in the late 1960s and early 1970s. The early experiments were aimed at producing a robust, fault-tolerant network that would have no single point of failure. The resulting ARPANET was the prototype for the development of the larger Internet, simultaneously serving as a testbed and linking network researchers. In 1985 the National Science Foundation (NSF) began to link its supercomputer centers via the Internet and in 1987 established NSFNET as a high-speed backbone for the Internet. Over time, and often in partnership with the private sector, numerous federal agencies began to support networking, moving what started as a defense technology into the general civilian research community, the higher-education community, and eventually into an expanding portion of the nation’s general population. In both computing and networking, federally funded research in partnership with private efforts has played a significant role in the development of technologies that have successfully migrated into the business sector and provided broad benefits for the whole nation. Architectural and Programming Issues The same underlying technology that advanced the microprocessor and made sophisticated networking possible also formed the roots of the High Performance Computing and Communications Initiative. It seemed clear to a few farsighted individuals in the computing community that as VLSI technology developed, it would favor computing structures that relied on replication of smaller computing units, as opposed to monolithic computers that relied primarily on very high speed circuits that were expensive to design, produce, maintain, and operate. This vision of high-performance computing brought two major technical challenges. Interconnection and memory architecture: how to unite large numbers of slower and cheaper processors or computers into systems capable of delivering truly high performance. Programming: how to program such collections of devices to solve large and complex problems. Architecture—Parallel computing. Generically, the approach of using multiple processors came to be called parallel computing or parallel processing. While the basic concept of parallel processing is not new, the development in the past 10 years of faster microprocessors and networks has led many in the computer community to believe that
OCR for page 4
some form of parallel computing holds the most promise for obtaining further improvements in computing performance. However, continued research and development will be needed to broaden the range of applications for which the actual speeds achieved by newer parallel designs exceed those of vector/parallel supercomputers in use today. The development of high-performance parallel computing has taken place in stages.5 The earliest attempts to build computers with a new structure used arrays of processors controlled centrally, with each individual processor doing the same operations on separate but parallel data elements. This approach, called single instruction multiple data (SIMD), proved useful only for limited applications. Later approaches focused on using groups of processors, again working in parallel, but with a richer form of control and coordination. This architecture came to be called multiple instruction multiple data (MIMD) because each of many processors concurrently executed its own set of instructions while accessing its own data stream. Today, MIMD systems are the dominant model and range in size from a few processors to hundreds of processors. The MIMD approach has demonstrated broad applicability than the earlier SIMD approach. Currently, most MIMD machines use mass-produced conventional microprocessors because of their low cost. One MIMD architecture came to be known as the massively parallel processor (MPP). MPPs commonly comprise collections of processors nodes co-located inside a common cabinet. Each processor node consists of a microprocessor, some memory, and some communication capability. Each node is much like the processing unit in a workstation. The MPP machines connect their processors with very high performance, specialized interconnections so as to permit very rapid exchange of interprocessor messages. Another MIMD architecture exploits groups of workstations connected by conventional local-area networks. These “workstation farms” have demonstrated some success on large computing problems requiring modest internode communication. Between the two extremes of the MPP and workstation farms lie a number of parallel architectures now being explored. Computer scientists, engineers, and manufacturers currently hold a variety of technical opinions about the best form of parallel architecture. The committee does not know how or when this spread of opinion about parallel architecture will converge. The outcome depends on the future of microprocessor design, the economics of microprocessor manufacture, and the techniques developed and used by the computer community to implement effective parallel computing software. Programming. The development of parallel computing represents a fundamental change not only in the machines themselves, but also in the way they are programmed and used. Sequential computers execute instructions in a single sequential stream; parallel machines can do many operations at once. To use fully the power of a parallel machine, a program must give the machine many independent operations to do simultaneously, and it must organize the communication among the processor nodes. Developing techniques for writing such programs is difficult and is now regarded by the committee as the central challenge of parallel computing. Individual programmers and users learning how best to program and utilize the new machines also face a difficult transition: they must learn to think about problems in new ways, programming the machines to do many things at once. Computer scientists are now developing new theoretical concepts and underpinnings, new programming languages, new algorithms, and new insights into the application of parallel computing. While much has been done, much remains to be done: even after knowledge about parallel
OCR for page 5
programming is better developed, many existing programs will need to be rewritten for the new systems. It is no surprise that progress in learning to take full advantage of the parallel technology has been slow; paradigm shifts are always slow, and this one carries a heavy legacy of past experience and existing programs and tool sets. It is clear, however, that the HPCCI has focused the attention of scientists from many disciplines, application developers, computer experts, engineers, and manufacturers on both the realized and potential benefits of parallel computing. Multidisciplinary teams have developed important parallel applications and will continue to do so. Manufacturers of computers utilizing more traditional architectures are developing new lines to take advantage of parallel computing. Thus, the new ideas developed with HPCCI support are being widely used throughout industry, academia, and government. In fact, it appears that parallel computing is poised for important applications in the storage and analysis of massive databases, as multimedia servers and delivery systems, and as powerful on-line transactions systems. Current Status While the preceding structural and technical information provides the larger context in which the HPCCI should be understood and evaluated, several current but narrower issues affecting the initiative also require comment. GAO and CBO Reports Recent studies of the HPCCI, one by the General Accounting Office (GAO) and the other by the Congressional Budget Office (CBO), have been regarded by some as being critical of the entire initiative. The committee received detailed briefings on these two studies and believes that it is important to offer another perspective on their focus and context. The report by the GAO6 did not attempt to evaluate the entire initiative, but focused instead on the research funding, the computer prototype acquisition activities, and the balance between hardware and software investments by the Advanced Research Projects Agency (ARPA). It recommended that ARPA (1) broaden its computer placement program by including a wider range of computer types, (2) establish and maintain a public database covering the agency HPCCI program and the performance characteristics of the machines it funds, and (3) emphasize and provide increased support for high-performance software. The report’s authors stated to the committee that while recommending improvements, they had found that ARPA had administered its program properly. Likewise, rather than attempting an overall evaluation of the initiative, the CBO report focused on the HPCCI’s role in promoting a new parallel-computer industry in the United States.7 It concluded that the private market demand for some HPCCI-developed technologies might be very limited and thus, that the potential for revenue generation by such an industry might be limited. It did not consider either the economic impact of the application of high-performance computing and communications by other sectors of the U.S. economy, or the potential cost of relinquishing national leadership in high-performance computing and communications. In keeping with CBO policy, it offered no recommendations.
OCR for page 6
Because each study’s focus is relatively narrow, neither should be used to make generalizations about the overall initiative. Recognizing, however, that these studies have been interpreted as broad indictments of the HPCCI, the committee will address in its final report the larger set of issues implied by the GAO and CBO reports, including the possibility of developing standards by which to judge overall HPCCI program performance. Reorganization of Science Policy Activities In early 1994, science policy activities in the executive branch were reorganized; broad oversight for the HPCCI is now provided by the National Science and Technology Council (NSTC) through its Committee on Information and Communication (CIC). The High Performance Computing, Communications, and Information Technology subcommittee and the National Coordination Office, both of which report to OSTP through the CIC, continue to promote ongoing cooperation among the involved HPCCI agencies. The NCO also continues as the external liaison and as the publisher of the annual “blue books.” The committee’s final report will examine the effects of the large number of committees associated with the HPCCI. The final report will also examine the recent reorganization’s effects on the NCO and on the initiative. Budget The HPCCI’s multiagency budget is more complex than it would be had the program been started from scratch within a single agency. While complexity is inherent in multiagency programs and budgets, it has added to the confusion about spending priorities and accomplishments for the initiative. When the High Performance Computing and Communications Initiative was proposed in the executive budget for FY 1992, the agencies involved identified from their existing FY 1991 activities a base that contributed to the goals of the program. According to Grand Challenges: High Performance Computing and Communications: The FY 1992 U.S. Research and Development Program, the “blue book” report accompanying the president’s budget, this initial base totaled $489.4 million.8 In each subsequent year, agencies have added to this base in two ways: (1) by identifying additional existing programs that contribute to HPCCI goals and (2) by reprogramming and relabeling agency funds to support relevant aspects of the HPCCI. To this base of “identified” activities, Congress has added some funding each year for new activities or the expansion of existing efforts. The result is that the $1.1 billion requested for fiscal 1995 is composed of three elements: (1) funds for the continuation of agency activities that were in existence when the HPCCI started and were designated in the 1992 base budget, (2) funds for existing or redirected programs that have since been designated as being a part of the HPCCI, and (3) additional funds for new activities or expansion of existing efforts. It is difficult to determine exactly how large each element is and to make interagency comparisons, because each agency has used slightly different approaches for identifying existing efforts and somewhat different formats for supporting program and budgetary detail. The committee will more closely examine the financial aspects of the HPCCI in its final report.
OCR for page 7
OBSERVATIONS An Effective Industry-Academia-Government Partnership The HPCCI builds on several decades of research, development, and technology transfer in computer science and engineering that have been marked by an effective partnership among industry, academia, and government. There is a substantial record of government-funded university research stimulating the development of new products and companies through the transfer of both technologies and trained people. Continued government investment in the core technologies of computer science and engineering will in all likelihood continue to yield major positive benefits to the nation’s economy. Contributions to a Paradigm Shift The HPCCI is contributing to a paradigm shift in high-performance computing from scalar and vector computing to a three-way mix that gives full weight to highly parallel computing. Parallel computing is becoming an established technology, and a variety of parallel architectures are proving to be useful for different computational tasks. Depending on a firm’s or agency’s mission, parallel computers can often be procured in the same manner as any established technology. However, a shortage of third-party-developed application software currently constrains the benefits being realized from parallel computing. While the committee cannot predict the precise future path of parallel architecture development, the overall importance of parallel computing is now widely recognized. Efforts under the HPCCI to foster its use are beginning to bear valuable intellectual fruit. In whatever direction the interplay of economics, technology, and software drives the revolution in parallel computing, the intellectual output attributable to the HPCCI can continue to provide the knowledge and personnel base to make parallel machines increasingly useful. Software and Algorithms Needed The development of software for highly parallel computers trails the evolution of the hardware. Thus, operating systems for parallel computers are immature, as are high-level language compilers. Few third-party commercial developers of major applications have as yet reworked their offerings to exploit highly parallel computing. Computer scientists and computational scientists in the user-disciplines do not yet know how to convert many applications to parallel computing. Discovery and development of new algorithms, as well as programming, will be crucial to obtaining maximum benefit from the paradigm shift to parallel computing. The committee will include a more detailed examination of algorithms and programming in its final report. Progress on Network Technology Impressive progress has been made in developing switches and transmission technology for high-speed networking. The speed of digital communications is rising rapidly, helped in part by high-speed network testbeds funded through the HPCCI.9
OCR for page 8
If one thinks of high-performance computing and communications as a tent, the thrust of the HPCCI so far has been to raise the center pole higher. It has given the nation more capable computers and faster communications, beginning the process of making many of these enhanced capabilities widely available. The objective of the National Information Infrastructure (NII) as a government initiative is to broaden further the base of the tent: the NII contemplates a network complex linking large numbers of users in schools, hospitals, businesses, homes, and government. Such a broad internetwork will inevitably raise technical challenges of a new and different kind, for example, economical interconnection of thousands of active users. Of course, the existing successes of the HPCCI can help with such problems. Beyond these, it will be necessary to provide a significant research and development component as part of the NII initiative. Cooperative Efforts on Broad Research Topics The HPCCI remains an interagency, cooperative effort among diverse federal agencies. While this aspect of the initiative may be the source of some confusion if one is looking for a single point of responsibility and program management, it also means that each federal agency has been able to pursue overall HPCCI goals within the structure of its own assigned mission. Centralized control would not only discourage agencies from supporting the initiative, but would also decrease the diversity of approaches being evaluated and thus reduce the scope of the HPCCI program. Thus, as a group, the HPCCI agencies have undertaken and supported valuable research across a wide spectrum of computer and communications topics. Information Gathering The National Coordination Office helps the mission agencies by channeling information from industry and the public to the agencies involved in the HPCCI. It has held meetings between representatives of industry and the HPCCI agencies to help assess both progress and future needs. However, the NCO’s information-gathering role is less well developed than its other functions. This is due partly to the absence of the anticipated Advisory Committee to the HPCCI and partly to the limitations of the Federal Advisory Committee Act on participation by people other than federal employees at government-sponsored meetings. NCO Support of Mission Agencies The National Coordination Office supports the mission agencies rather than directing them. By avoiding actual direction the NCO leaves mission judgments in the hands of responsible agency officials who are accountable for the allocation of their resources. By avoiding the appearance of direction the NCO encourages an appropriate diversity of research projects as each agency capitalizes on its best ideas. Mission agencies cooperate effectively with each other and with the NCO precisely because it does not threaten their autonomy. This cooperation could easily vanish were the NCO seen as functioning with too heavy a hand. The committee believes that the value of interagency cooperation outweighs the potential benefits that might be gained through more centralized management of the HPCCI.
OCR for page 9
Recently, the dimensions of the need for clear communication about the HPCCI have become apparent: congressional oversight activities and other indicators suggest that the HPCCI is of concern to a growing constituency, and that often it needs to be explained in detail to a variety of audiences. Such an effort will add substantially to the work of the NCO, which is currently headed by a half-time, permanent-position director who holds a concurrent, half-time appointment as director of the National Library of Medicine. The other NCO staff positions are a mix of permanent positions, contract positions, and temporary positions filled by individuals on loan from other federal agencies for limited periods of time, generally no more than one year. While the NCO reports to OSTP on programmatic matters, administrative functions such as office space, salaries, and benefits are handled through the National Institutes of Health. The temporary nature of some of the NCO positions jeopardizes continuity, cumulative insight, and intellectual ownership. Further, limited staff time and resources raise questions about the NCO’s capacity to meet the challenge of the growing volume, complexity, and urgency of the outreach efforts needed for the initiative. Advisory Committee for HPCCI The intent of Congress to have active nongovernmental input into the HPCCI is reflected in its mandate for an Advisory Committee for the initiative. Because the HPCCI encompasses not only scientific objectives but also enhancement of the industrial competitiveness of the United States, an Advisory Committee with strong application-industry, computer industry, academic, and government representation is necessary to ensure the increasing relevance of the HPCCI to industry needs. RECOMMENDATIONS A Stronger National Coordination Office Recommendation 1: Because the size, maturity, and visibility of the High Performance Computing and Communications Initiative now warrant a stronger National Coordination Office (NCO) than was originally envisioned, the committee recommends the following approach: The National Coordination Office should provide an increasingly comprehensive and intensive outreach and education service. It should continue to serve a coordination function within the executive branch, organizing interagency meetings and fostering interagency cooperation. It is appropriate for the NCO to consolidate HPCCI budget information across agencies and to aid the agencies in assessing their budget requests with respect to their program commitments. However, budgetary authority should remain in the individual agencies to preclude interference with their missions. To continue the role outlined above, the NCO needs a full-time director or coordinator. Having a part-time director has served well to this point, but the broadening of the HPCCI demands leadership unencumbered by other major responsibilities. The NCO should remain within the Office of Science and Technology Policy structure.
OCR for page 10
Immediate Appointment of HPCCI Advisory Committee Recommendation 2: The committee encourages the immediate appointment of the congressionally mandated Advisory Committee intended to provide broad-based, active input to the HPCCI. The HPCCI could be improved by input from and review by an Advisory Committee with balanced representation from industry, academia, and government, especially current and potential users of high-performance computing and communications. If appointment of such a committee is not feasible, some alternate mechanism should soon be developed to provide similar input. NOTES 1. Committee on Physical, Mathematical, and Engineering Sciences; Federal Coordinating Council for Science, Engineering, and Technology; Office of Science and Technology Policy. 1993. High Performance Computing & Communications: Toward a National Information Infrastructure, 1994. Washington, D.C.: Office of Science and Technology Policy, p. 1. 2. Each year the director of the Office of Science and Technology Policy submits a report on the HPCCI to accompany the president’s budget. It describes prior accomplishments and the funding and activities for the coming fiscal year. These reports have become known as “blue books” after the color of their covers. 3. NSF Blue Ribbon Panel on High Performance Computing. 1993. From Desktop to Teraflop: Exploiting the U.S. Lead in High Performance Computing. Washington, D.C.: National Science Foundation , pp. 1–5. See also Computer Science and Telecommunications Board. 1994. Information Technology in the Service Society: A Twenty-First Century Lever. Washington, D.C.: National Academy Press. 4. Patterson, David A., and John L.Hennessy. 1994. Computer Organization and Design: The Hardware/Software Interface. San Mateo, Calif.: Morgan Kaufmann Publishers, p. 21. 5. Computer Science and Technology Board. 1988. Global Trends in Computer Technology and Their Impact on Export Control. Washington, D.C.: National Academy Press. 6. U.S. General Accounting Office, Information Management and Technology Division. 1993. High Performance Computing: Advanced Research Projects Agency Should Do More to Foster Program Goals. Washington, D.C.: U.S. General Accounting Office, pp. 2–5. 7. Congressional Budget Office. 1993. Promoting High-Performance Computing and Communications. Washington, D.C.: U.S. Government Printing Office, p. iii and p. 1. 8. Committee on Physical, Mathematical, and Engineering Sciences; Federal Coordinating Council for Science, Engineering, and Technology; Office of Science and Technology Policy. 1991. Grand Challenges: High Performance Computing and Communications: The FY 1992 U.S. Research and Development Program. Washington, D.C.: Office of Science and Technology Policy, p. 24. The agencies that had activities included in the FY 1991 base were the (Defense) Advanced Research Projects Agency, Department of Energy, National Aeronautics and Space Administration, National Science Foundation, National Institute of Standards and Technology, National Oceanic and Atmospheric Administration, Environmental Protection Agency, and National Institutes of Health/National Library of Medicine. 9. Computer Science and Telecommunications Board. 1994. Realizing the Information Future. Washington, D.C.: National Academy Press.
Representative terms from entire chapter: