Page 92

B
High-Performance Communications Technology and Infrastructure

HIGH-PERFORMANCE COMMUNICATIONS TECHNOLOGY AND INFRASTRUCTURE ADVANCE

The performance/cost imperatives of communications have driven the technology from parallel to serial, from hundreds of slower wires to a few very fast fibers. Fiber-optic transmission offers stunning bandwidths: 100,000 telephone calls or 800 video channels on one pair of fibers. Communications is different from computing. Value often comes from whom one can talk to, rather than how rapidly. Issues of scaling are very important. The scale of the needed networking raises a host of new research issues as to how millions of users can attach to the network.

High-performance computing and high-performance communications support each other in complex ways. Communications has become digital, and the switching of fast digital signals requires high-performance computing technology. On the other hand, very fast computer-to-computer communications are crucial for many applications. Today 16 percent of investment in the High Performance Computing and Communications Initiative (HPCCI) is directed at communications. The communications content of the HPCCI has two aspects: research and development to advance communications and related capabilities (see Table B.1), and delivery of access to communications-based infrastructure to researchers to facilitate their work (see Table B.2). The introduction of the fifth HPCCI component, Information Infrastructure Technology and Applications (IITA), in 1993 appears to extend the second aspect: awards and activities associated with this component appear to emphasize making existing capabilities more useful and more widely used, as opposed to developing new communications-based capabilities, which appears to be largely supported under the National Research and Education Network (NREN) component of the HPCCI.

The Internet is the centerpiece of the present HPCCI communications infrastructure program; it includes the network elements (backbone, regional, and ''connections") supported under the NREN program. Thanks to federal support of internetworking technologies that have been applied in the Internet generally and specifically in NREN-supported elements (NSFNET, ESnet, and the NASA Science Internet), the United States has a strong lead in these technologies worldwide. The United States is home to a vital industry that supplies related equipment and software, including businesses begun as spinoffs from academic research activity. See Box B.1 for an example of how government, academic, and industrial investments can complement each other to accelerate the development of a key communications technology.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



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 92
Page 92 B High-Performance Communications Technology and Infrastructure HIGH-PERFORMANCE COMMUNICATIONS TECHNOLOGY AND INFRASTRUCTURE ADVANCE The performance/cost imperatives of communications have driven the technology from parallel to serial, from hundreds of slower wires to a few very fast fibers. Fiber-optic transmission offers stunning bandwidths: 100,000 telephone calls or 800 video channels on one pair of fibers. Communications is different from computing. Value often comes from whom one can talk to, rather than how rapidly. Issues of scaling are very important. The scale of the needed networking raises a host of new research issues as to how millions of users can attach to the network. High-performance computing and high-performance communications support each other in complex ways. Communications has become digital, and the switching of fast digital signals requires high-performance computing technology. On the other hand, very fast computer-to-computer communications are crucial for many applications. Today 16 percent of investment in the High Performance Computing and Communications Initiative (HPCCI) is directed at communications. The communications content of the HPCCI has two aspects: research and development to advance communications and related capabilities (see Table B.1), and delivery of access to communications-based infrastructure to researchers to facilitate their work (see Table B.2). The introduction of the fifth HPCCI component, Information Infrastructure Technology and Applications (IITA), in 1993 appears to extend the second aspect: awards and activities associated with this component appear to emphasize making existing capabilities more useful and more widely used, as opposed to developing new communications-based capabilities, which appears to be largely supported under the National Research and Education Network (NREN) component of the HPCCI. The Internet is the centerpiece of the present HPCCI communications infrastructure program; it includes the network elements (backbone, regional, and ''connections") supported under the NREN program. Thanks to federal support of internetworking technologies that have been applied in the Internet generally and specifically in NREN-supported elements (NSFNET, ESnet, and the NASA Science Internet), the United States has a strong lead in these technologies worldwide. The United States is home to a vital industry that supplies related equipment and software, including businesses begun as spinoffs from academic research activity. See Box B.1 for an example of how government, academic, and industrial investments can complement each other to accelerate the development of a key communications technology.

OCR for page 92
Page 93 TABLE B.1 HPCCI Program Activities in Communications Research, FY 1995 ComponentA AgencyB Funding Request for FY 1995 (millions of dollars) Activity NREN ARPA 43.10 Networking IITA ARPA 23.00 Global grid communications BRHR NSF 11.30 Very high speed networks and optical systems NREN NSA 03.50 Very high speed networking NREN NSA 02.60 High-speed data protection electronics NREN DOE 02.00 Gigabit research and development NREN NIST 01.75 Metrology to support mobile and fixed-base communications networks ANREN, National Research and Education Network; IITA, Information Infrastructure Technology and Applications; BRHR, Basic Research and Human Resources. BARPA, Advanced Research Projects Agency; NSF, National Science Foundation; NSA, National Security Agency; DOE, Department of Energy; NIST, National Institute of Standards and Technology. TABLE B.2 HPCCI Program Activities in Communications Infrastructure, FY 1995 ComponentA AgencyB Funding Request for FY 1995 (millions of dollars) ActivityC NREN NSF 46.16 NSFNET NREN DOE 14.80 Energy sciences network (ESnet) NREN   12.70 NREN NREN NOAA 08.70 Networking connectivity NREN NIH 06.50 NLM medical connections program NREN NIST 02.20 NREN deployment and performance measures for gigabit nets and massively parallel processor systems IITA NIH 02.00 NCI high-speed networking and distributed conferencing NREN EPA 00.70 State network connectivity ANREN, National Research and Education Network; IITA, Information Infrastructure Technology and Applications. BNSF, National Science Foundation; DOE, Department of Energy; NOAA, National Oceanic and Atmospheric Administration; NIH, National Institutes of Health; NIST, National Institute of Standards and Technology; EPA, Environmental Protection Agency. CNLM, National Library of Medicine; NCI, National Cancer Institute.

OCR for page 92
Page 94 BOX B.1 Federal Government Participation in the Development of Asynchronous Transfer Mode The federal government, through the HPCCI and other programs, played a part in the development of the asynchronous transfer mode, or ATM, standard, and most importantly, played a very significant role in developing broad support for ATM as an important switching technology for high-speed computer networks. ATM was developed as a switching technology by the telecommunications community for application in the so-called broadband integrated services digital network, or BISDN. However, its development did not occur without some controversy. Telecommunications switching had always been based on circuit-switching technologies, which are very well suited to providing fixed-rate connections in a network. Multirate circuit switching can be used to provide connections that are any multiple of some basic rate. Narrowband ISDN is based on these technologies, and most switching and transmission experts in the telecommunications industry expected a straightforward extension of ISDN circuit-switching technology, based on multiples of a 64-Kbps basic rate, into the broadband realm of speeds in the 1 55-Mbps range. Variable rate communications services were clearly needed for data communications, as demonstrated by the ARPANET in the 1970s. However, it was also recognized that video and even voice services might be provided in a more efficient manner than was possible through circuit switching if the basic network service could support variable rate communications. Between 1984 and 1987 researchers in leading telecommunications and academic laboratories developed a number of fast-packet switching, or FPS, systems that laid the technological groundwork for the ATM standard. A noted academic researcher participating in this effort was Jonathan Turner, whose work was supported by both NSF and industry funding. Telecommunications industry laboratories such as AT&T, Bellcore, CNET (France), and Bell Telephone Manufacturing (Belgium) also played a significant role. Bellcore, for example, built the first prototype broadband central office in late 1986, and it contained a packet switch that operated at 50 Mbps per line. By 1987, this rate had been extended to nearly 1 55 Mbps per line. These efforts, in essence, proved that packet switching was capable of running at the data rates required by BISDN. At this time, ARPANET and NSFNET were running at speeds of only 64 Kbps to 1.5 Mbps, and local area networks such as Ethernet ran at speeds of only 10 Mbps. In late 1986, Gordon Bell, then at NSF, visited Bellcore and received a briefing on fast CMOS-based packet switching technology that had been simulated at speeds of 150 Mbps. Several months later he called a workshop of computer networking researchers to discuss the future of high-speed computer networking. Out of that workshop, from David Farber (University of Pennsylvania) and Bob Kahn (Corporation for National Research Initiatives), came a mid-1987 proposal to NSF to form the gigabit testbed projects. The telecommunications industry was quite active during this time in developing a standard based on packet switching. In 1988, the Consultative Committee on International Telephony and Telegraphy selected (and named) ATM as the standard for BISDN switching. This action served to focus the telecommunications industry on the development of this technology. However, in 1 988 ATM was little more than a name, a basic packet format, and a common cause across the telecommunications industry alone. In 1989, the gigabit testbed projects were formally initiated with (D)ARPA and NSF funding. One of the objectives of the gigabit testbeds was to determine what switching technologies were appropriate for use in gigabit networking. There were several major contenders: ATM, HiPPi (a switching technology favored by the supercomputer community), and PTM (a proprietary packet-switching technology advocated by IBM, different from ATM primarily in that it used variable-sized packets). Although it would be several years before conclusive results from the testbeds were produced, the use of ATM in the testbeds gave it wide exposure within the federal government's technology community. continues

OCR for page 92
Page 95 BOX B. 1—continued In 1990, the first company devoted solely to producting ATM products was formed. The founders of FORE Systems were first exposed to ATM technology through their participation in the NECTAR gigabit testbed. Cross-testbed exchanges were encouraged by the gigabit testbed program, and in 1989 and 1990 there were meetings of the Bellcore and MIT researchers working on ATM for Aurora and the Carnegie Mellon University researchers working on NECTAR. FORE was founded in 1990 and subsisted through its first product development cycle largely on research and prototyping contracts from ARPA and the Naval Research Laboratory. FORE's first products were delivered to the Naval Research Laboratory in 1991. FORE today is a rapidly growing company with nearly 300 employees and one of the leading manufacturers of ATM switching equipment. In 1990, a collaboration was formed between Apple, Bellcore, Sun Microsystems, and Xerox to attempt to gain the acceptance of ATM as a new local area networking standard. This group, which eventually expanded to include Digital and Hewlett-Packard, produced the first published specification of ATM for Local Networking Applications in April 1991. ARPA participated in informal discussions with this group via its ARPA Networking Principal Investigator meetings, looking for opportunities to fund technical work that would be necessary to make ATM useful for local area networking. One notable result of ARPA's funding was the first implementation of the Q.2931 ATM signaling standard, which was required to allow ATM to implement switched (as opposed to permanent) virtual circuits. Another result was the implementation, by Xerox, of an ATM switch architecture that is well suited for low-cost implementation. This work, funded in part by ARPA, served to accelerate the development of an industry standard for local ATM by the ATM Forum. Once ATM had been established as both a computer and telecommunications networking standard, the role of the federal government turned toward developing an early market for high-speed networking equipment and services. Several notable programs greatly expanded this early market. One has been the gigabit testbeds. Some of the first installations of 2.4-Gbps SONET and ATM by several telecommunications carriers were for these testbeds. Although the federal government did not spend a single dollar for these facilities, its leadership of the gigabit testbed program was critical in getting telecommunications carriers to construct these very expensive facilities for the testbeds. Telecommunications equipment manufacturers benefited directly and saw their development of very high speed (full-duplex 622-Mbps) ATM equipment greatly accelerated. Some direct federal procurements of ATM and SONET services are still playing a key role in the development of the marketplace. An important procurement of ATM and SONET services and technology was the Washington Area Bitway (WABitway), the first significant sale of high-speed SONET services by Bell Atlantic. In early 1994, NSF announced the selection of a number of companies to provide ATM-based communications services for the new NSFNET. Telecommunications companies involved in this project include Ameritech, MCI, MFS Datanet, Pacific Bell, and Sprint. Eight HPCCI program activities are directed primarily at communications infrastructure, principally supporting deployment of the Internet within each agency's community. Several of the infrastructure programs are building on early results of the gigabit testbed research. For example, ATM and SONET networking technologies, first deployed in the gigabit testbeds in 1992, appear in some form in many of the FY 1995 infrastructure activities. Developing and broadening access to information infrastructure pose many research issues. Information infrastructure is more complex than networks, per se, and the computing and communications research community has already helped to explore and define fundamental concepts, for example, the concept of "middleware" to cover the kind of internal services that help to transform a network into information infrastructure. Research into how to implement such services has begun under the HPCCI umbrella. More specifically, the National Aeronautics and Space Administration, the National Science Foundation, and the Advanced Research Projects Agency have combined to fund research to support the development of digital libraries, providing a vehicle for exploring many concepts associated with information infrastructure (NSF, 1993).

OCR for page 92
Page 96 The concept of a better nationwide information infrastructure, itself connected to a global information infrastructure, poses yet other concerns associated with interconnecting multiple kinds of networks from multiple kinds of providers to multiple kinds of users offering multiple kinds of services. This construct adds great complexity, increasing the emphasis on scale and architecture and adding in such concerns as heterogeneity of systems, decentralization of control and management, routing, security, and so on. There have been many government, academic, and industry studies under way to identify and clarify these research issues. Although the newspapers are filled with announcements of corporate alliances, new venture formations, and new product introductions more or less linked to the advancement of the nation's information infrastructure, significant advances call for the solution of many technical problems and therefore for a significant research effort.