Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Report of the Pane! on the Network Systems and Communications Industry The Panel on Network Systems and Communications, one of five panels formed by the Committee on the Impact of Academic Research on Industrial Performance, was asked to examine the impact of academic research on the performance of the network systems and communications industry and recom- mend ways based on trends in the industry and the research community to in- crease this impact. The panel of six included three members of NAB (all from industry), one other member from industry, and two from academia. Three of the panel members were also members of the parent committee. The panel reviewed the literature, developed several case studies, and sent a questionnaire to experts in academia, the computer industry, the network systems and communications industry, and government. The questionnaire was followed by a workshop at- tended by approximately 30 senior individuals in the network systems and com- munications sector (see Addendum). The network systems and communications business sector flourished throughout the l990s, when the growth of the Internet, the technologies that implement it, and the businesses and services that use it were unprecedented. Telecommunications services especially wireless digital telephones and paging services also grew rapidly. Much of this success was attributable to exponential improvements in the performance-to-cost ratio of microelectronics over the past three decades. Technical innovations emerging from within the industry and from academic research have been essential. Some innovations were the culmination of decades of research; some were short-term developments that entered the 29
30 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE market via start-up companies; and some were incremental improvements to existing products. In the last 30 years, digital technologies have transformed the telephone network from an analog system to a computer-controlled system with digital switching and transmission. The process of digitalization has changed the indus- try from two distinct businesses computers and communications to one busi- ness in which computers and communications are intermingled in products and services. This convergence was accelerated by advances in microelectronics and increases in the bandwidth available for communications (Messerschmitt, 1996~. The result is increasingly pervasive data networking, based largely on the packet- switching technologies that emerged from academic and industrial research to spawn the Internet. The network systems and communications industry has a large and expand- ing services component. For the telecommunications industry, which has always been a service provider, the challenge is to invent and offer customers new, valuable services that generate new sources of revenue. For the computer portion of the industry, high-performance communications are making a wide range of new services feasible. Examples include remote sensors and control systems; integrated supply chain management systems; application service providers; full- time, real-time stock quotes; and instant messaging. DEFINITION OF THE INDUSTRY The network systems and communications industry must be defined very broadly. It clearly includes the manufacturing of telecommunications equipment and the services that use such equipment, such as telephony, wireless telephony, broadcast television, cable and satellite television, radio, and Internet service. Both the equipment and services sectors increasingly require computing equip- ment and software, and, in fact, the computer and communications industries are no longer separate industries. For example, cellular telephony depends on a broad range of technologies: the cell phone contains a liquid crystal display, an embed- ded computer with a lot of software, and advanced chips that integrate most of the components of a high-frequency radio; the transmission formats depend on ad- vances in digital speech compression, signal modulation, and coding; the base stations depend heavily on digital integrated circuits and computers for switching and control and fiber-optic links between them; tracking a moving telephone requires that computers at adjacent base stations exchange protocol messages for the handoff; and the billing, provisioning, and maintenance of the service require large-scale computing and software systems of the service provider. Separating this integrated system into "communications" and "computing" components is simply not possible. In short, computing and communications equipment and services have converged, creating new business and technical opportunities.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 31 The explosive growth of the Internet is the most visible manifestation of this trend toward convergence. The technologies underlying the Internet- just like those that underlie the cellular telephone include computing and communications. Special computers serve as routers, and network services knit together the transmission links and implement the collection of Internet protocols that carry Internet traffic. The explosive growth of the Internet, however, is attributable not to these basic provisions which existed before 1993 but to new services that created consumer demand: electronic mail. the World Wide Web of information and its associated browser software; chat groups; real-time delivery of audio and video media; online merchandising; banking and financial transactions; supply-chain integration of suppliers and customers; and numerous other applications. Some applications merely ex- tend existing internal information technology systems to provide Internet access. But others, such as eBay's success with online auctions, are entirely new business concepts. As the Internet becomes more pervasive, old ways of computing, in which data was created, stored, and manipulated at a single site, are giving way to networked systems in which data can be accessed remotely and shared extensively. The computers embedded in everyday objects telephones, cars, televisions, furnaces, hi-fi equipment are becoming increasingly capable and increasingly networked. Some cars already can connect with a diagnostic and help center by cellular telephone or satellite communications. Home networks in which multiple personal computers in a household are linked over existing telephone wires and short-range wireless devices will soon make networking of appliances routine. A world in which all devices have an Internet address is not out of reach. Thus computers increasingly require communications to fulfill their functions, and communications increasingly require computers to fulfill theirs. The technologies of computing and communications are becoming indis- tinguishable. All of them depend on software to express functions at all levels in the network. A few years ago, a modem was a complex, integrated circuit. Today, with more complex algorithms and faster computers, modems are writ- ten in software embedded within digital signal processors. Many algorithms can be used equally well in computing and communications settings. For ex- ample, schemes to digitize and compress video signals are useful both for manipulating and storing video information on a computer disk and for trans- mitting it over digital communications channels. Similarly, encryption technol- ogy can be used to protect sensitive information in a computer system or in transit over a network. These three trends convergence, embedding, and network applications- characterize the network systems and communications sector. The panel's assess- ment of the contributions of academic research to this industry is based on this broad definition.
32 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE TABLE 2-1 Sales and Employment in the Information Technology Industry, 2000 Sales Number NAICSa Revenues of Jobs Code ($ billions) (1,000) IT Manufacturing Computer and peripheral equipment 3341 $110.0 190 Communications equipment 3342 119.3 291 Software 5112 88.6 331 Semiconductors and other electronic components 3344 168.5 621 IT Services Data processing services 5142 42.9 296 Telecommunications services 5133 354.2 1,165 aNorth American Industrial Classification System. Source: U.S. Bureau of the Census, 2002. Size Because our definition has vague boundaries and because the industrial classifications used to gather statistics have not been adapted to the rapid changes in the industry, it is difficult to determine the size of the network systems and communications sector. Table 2-1 summarizes sales and employment in the information technology industry based on Bureau of the Census data (U.S. Bu- reau of the Census, 2002~. Taken together, sales of computer and communica- tions equipment and services (all information technology minus semiconduc- tors) were about $715 billion in 2000, and the industry employed more than 2.2 million people (U.S. Bureau of the Census, 2002~. Expenditures for information-processing equipment increased almost 10 percent per year on aver- age from 1970 to 1994; the corresponding figure for computers and peripherals was 27.5 percent (NRC, 1999~. A 1999 survey found that telecommunications manufacturing was growing by 16.3 percent annually, computer software by 16.6 percent, and computer hardware by 9.5 percent (CTIS, 1999), however these rates have dropped significantly since early 2000. Structure The role of research and innovation in the network systems and communications sector can best be understood in the context of the structure of the industry, which influences the mechanisms of innovation and thus how new technologies and prod- ucts are introduced. The very general description that follows is intended only to reveal similarities and differences with the other industries studied in this report. Manufacturing The structure of the computer industry is horizontal; the communications industry was vertically integrated but has been rapidly changing to a horizontal
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 33 structure as well. In a horizontal structure, numerous suppliers manufacture parts and components that many integrators assemble into subassemblies that are then assembled into final products by numerous competing original equipment manu- facturers. The multiplicity of companies at each manufacturing step ensures in- tense competition throughout the production process, not only in terms of price but also on a wide rage of performance characteristics. For example, manufac- turers of personal computers buy disk drives from any one of about a dozen suppliers. A company that needs a customized integrated circuit may design the circuit but use one of several competing semiconductor foundries to manufacture it. Specialized circuit board assembly firms can assemble and test complete cir- cuit boards, giving an electronic design firm the ability to design and sell a unique computer interface board with custom chips without having to invest in either chip or board manufacturing facilities. The divestiture of AT&T and the subsequent deregulation of communica- tions services forced the communications industry to change from a vertical to a horizontal structure. Today there are many vendors of telecommunications equip- ment and components. Custom integrated circuits can implement very complex communications functions; coupled with custom-built and proprietary software designs, equipment vendors compete intensely in terms of technology, reliability, and cost of ownership. Another important feature of the network systems and communications sec- tor is its reliance on components with well defined interfaces. Integrated circuits are a good example: the physical, electrical, and logical behaviors of chip inter- faces are specified by the manufacturer and used by the customer to determine how to incorporate a chip into a subsystem with other components. Subsystems then become components of still larger systems. Software, another component, plugs into the operating system that supports it by linking the software interfaces (sometimes called application programming interfaces, or APIs). A piece of soft- ware that is compatible with a certain operating system adheres to the interfaces provided by the operating system. Computer systems are built from complex hierarchical assemblies of subsystems and components, sometimes hundreds or even thousands of them. Some of the components are custom built, and some are standard. Thus, interfaces give rise to components, which in turn give rise to businesses structured around buying and selling components. Key component interfaces become industry standards, which are usually adopted by industry groups to hasten the spread of a new technology, increase sales volume, and, therefore, decrease cost. Standards maintained by a group with broad representation from competitive suppliers are said to be open standards. For example, the Personal Computer Memory Card International Association is an industry group that establishes standards for interchangeable interface cards for laptop computers. The standards group includes several producers of cards and several producers of laptops to ensure that the standard cannot be manipu- lated to benefit one competitor over another. By contrast, standards promulgated
34 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE by a single vendor are said to be proprietary. For example, the programming interface for Microsoft's Windows operating system is proprietary; Microsoft specifies it and can change it at will. Standards play a special role in communications. Broadly speaking, they are nec- essary to ensure that components and subsystems connected via a communications channel can operate together (e.g., they obey the same conventions for encoding voice signals, multiplexing many simultaneous phone calls on a single channel, performing operation and maintenance functions). Standards of this kind are necessary for guaran- teed, sustained interoperability, and changes must be carefully designed to avoid even slight interruptions of network service. New versions of standards must be designed so they can be introduced incrementally, connect new equipment to old, test new proto- cols, and so on. The same considerations apply to Internet protocols. Services Communication services (e.g., voice and data transmission, switching, and distribution) are a major portion of the network systems and communications in- dustry. The number and structure of telecommunications service providers have been in constant flux since the divestiture of AT&T and the deregulation of local telephone services. First, new companies emerged offering wireless telephone ser- vices. Then another group of new companies emerged as Internet service providers. To increase their revenue, carriers have been developing value-added services, such as voice mail, call forwarding, call waiting, 800 service, electronic mail, and virtual private networking, along with conventional transmission and switching services. Internet service providers provide national and regional portals that offer news, chat rooms, advertising, and direct access to the World Wide Web. Computing services are also a major element of the industry. System integration, the design and deployment of communications and information systems for large clients, has become a major source of revenue for many equipment vendors. In recent years, an important service has been to implement network capabilities across compa- nies' existing computer systems. In some cases, networking has focused on providing Internet access for employees and customers; in others, the focus has been on the development of internal networks linking production and distribution facilities across the company. So far, neither academic nor industrial research has addressed the problems of service delivery in a structured and sustained manner. INNOVATION SYSTEM Most innovations are incremental improvements, such as design refinements, improvements in technology and manufacturing processes, a better understanding of customer needs, and integration of previously separate products. For example, impor- tant performance metrics for communications equipment include low power and high density (so that many circuits can be accommodated in the confined spaces of wiring
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 35 closets, boxes mounted on telephone poles, and even central switching offices). Both power and density can be improved by advances in integrated-circuit technologies, which in turn, derive primarily from incremental improvements in fabrication equip- ment, processing steps, and materials. Research results may be the basis for some of these improvements, and research has achieved major breakthroughs in these areas; this research is performed or funded by materials, equipment, and microchip fabrica- tors, not by the telecommunications equipment manufacturers (see Box 2-1~.
36 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE For many businesses, vendors of materials, products, and services throughout the supply chain are major sources of innovation. Buying an integrated-circuit chip, for example, implicitly buys a share of the dramatic improvements in price and performance of integrated circuits (Moore's Law).i Over time, innovations will make the chip faster or cheaper or more capable. A telecommunications carrier that wants to deploy Synchronous Optical Network (SONET), a transmission protocol that defines optical carrier levels and their electrically equivalent synchronous transport signals, can purchase switches, multiplexers, and test equipment from the vendors who developed SONET technology. This pattern is a direct consequence of the "horizontal" structure of the industry. Dell Computer, for example, does not have in-house R&D; in effect, Dell is a broker that negotiates attractive deals to buy components and computer-assembly services for its build-a-computer-to-order busi- ness. Dell depends on R&D investments by its vendors, especially Intel and Microsoft, that make the microprocessors and operating system software on which the personal computer business depends. Dell's innovations have been in its busi- ness model and supply-chain management, not in its technology. Innovation can also be purchased by acquiring other companies, especially venture-capital-backed start-up companies that have introduced new products with new technologies. A start-up company is a new business, often with an innovative technology but with considerable risk. Often the innovative technol- ogy has its origins in academic research. If the company makes good progress, both in technology and in the market (e.g., beta testing, or success in getting its approach adopted by standards consortia), it becomes an attractive target for a larger company seeking to strengthen its technology or product line. For ex- ample, Texas Instruments bought Amati; Fore Systems bought Berkeley Net- works and Marconi bought Fore Systems; Cisco bought Granite Systems; and Broadcom bought Epigram. Each of the acquired companies had ventured into a new technical area. Epigram, for example, had devised a way to use home tele- phone wiring to transmit 10Mb Ethernet traffic and had made progress in stan- dardizing the scheme through the Home Phoneline Networking Alliance. Broadcom, itself an innovative fabless chip company specializing in integrating analog and digital functions of cable and twisted-wire modems, saw buying Epigram as a natural way to enhance its core business. Although high-tech start-ups seldom do research in the classic sense, many behave much like "applied research" projects in an industrial laboratory. They formulate technically aggressive plans based on established principles to pursue and evaluate; the results of experiments often inform several products. For ex- ample, Transwitch attempted to increase the telecommunications protocol process- ing integrated on a single chip, as well as to partition the chip functions into an "architecture" so that a small number of chip designs could be used to build a wide variety of telecommunications products. Both Amati and Epigram conceived ways of using advanced signal-processing techniques to adapt digital transmission to the characteristics of real-world, twisted-pair copper wires (Amati) or in-house
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 37 telephone wiring (Epigram). The technology-development activities of these com- panies are much like those in industrial research, but they are done in a commercial setting and with strong incentives to bring innovations to market rapidly. Industrial research is concentrated in the laboratories of a few of the largest companies, such as Intel, Microsoft, IBM, Compaq, Lucent, AT&T, Hewlett- Packard, Sun Microsystems, and Xerox. Although many of these firms invest 10 to 15 percent or more of revenues in R&D each year, the vast majority of this is for "development," that is, for the engineering of the next generation of products. Research focused on objectives more than 18 months or one or two product cycles out is estimated to be, at most, 5 percent of that 10 to 15 percent, or far less than 1 percent of revenue.2 A few large companies eschew research, preferring instead to buy innovative companies (e.g., Cisco). Companies in the services sector, how- ever, generally do not engage in or support research. For example, at MCI, which is generally considered a technology leader, the advanced technology group is primar- ily concerned with testing new equipment and working with vendors to solve interoperability and operation, administration, and management problems. Industry research is usually driven by market needs but often includes some fundamental or long-range projects as well. For example, IBM's research on the Internet and electronic commerce includes some long-term work on cryptographic systems for security and authentication. Industrial research often links advanced technologies to emerging product needs. For example, as the Java programming language became popular, industry laboratories at Sun, IBM, and elsewhere launched projects to devise advanced techniques for the compilation, synchroni- zation, and code simplification required for its implementation. Previous research results in these areas had not adequately addressed the needs of the Java lan- guage, of today's large memories, or of multiprocessor servers. Some of this research is fundamental in the sense that it can be applied to problems other than Java language implementations. In fact, even though research in engineering fields is usually targeted toward meeting specific engineering needs, the results are often useful for many other applications. One of the companies' aims in operating research laboratories is to expand their capability for bringing in new ideas and new people (Cohen and Levinthal, 1990~. The laboratory is expected to recruit people who cannot be recruited by an engineering organization; it is also expected to interact with the intellectual community by attending conferences, publishing papers, collaborating with uni- versities, or entering partnerships with other companies; and it is intended to counteract the risk inherent in the narrow focus of engineering projects on prod- uct development. A Culture of Innovation Innovation in the network systems and communications industry can take many paths. Even when research plays an essential role, there is no linear path
38 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE 1965 1970 1975 1980 1985 1990 1995 2005 Timesharing Client~server Commuting e ; Graphics Interne1 LANs. Workstations Graphical user interfaces VLSI design I' RISC processors . , . 1965 1970 1975 | University ~ Industry R&D CTSS, Multics / BED Unix SDS 940, 360/67, VMS Berkeley, CMU, CERN PARC, DEC, IBM Novell, EMC, Sun, Oracle Sketchpad, Utah GM/IBM, Xerox, Microsoft E&S, SGI, ATI, Adobe Spacewar (MIT), Trek (Rochester) Atari, Nintendo, SGI, Pixar ARPANET, Aloha, Internet Pup DECnet, TCP/IP Rings, Hubnet Ethernet, Datakit, Autonet LANs, switched Ethernet Lisp machine, Stanford Xerox Alto ( to World Wide Web 1980 1985 1990 1995 2005 Xerox Star, Apollo, Sun Engelbart / Rochester Alto, Smalltalk Star, Mac, Microsoft Berkeley, Caltech, MOSIS many Berkeley, Stanford IBM 801 SUN, SGI, IBM, HP FIGURE 2-1 Examples of academic government-sponsored (and some industry- sponsored) IT research and development in the creation of commercial products and industries. Source: NRC, 2003.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 1965 1970 1975 1980 1985 1990 1995 2005 Relational databases . ~ from/Tternet Parallel databases Datamining Parallel computing RAID/disk servers Portable com inundation I.... 1 ~2 \-- World Wide Web . V, . . , Speech recognition Broadband in last mile 1965 1970 1975 \\ · ~ ~ ~ ~ - 1980 1985 1990 1995 2005 Berkeley, Wisconsin IBM Oracle, IBM, Sybase Tokyo, Wisconsin, UCLA IBM, ICE JCL, Teradata, Tandem Wisconsin, Stanford IBM, Arbor IRI, Arbor, Plato Illiac 4, CMU, Caltech, HPC IBM, Intel CM-5, Teradata, C ray T3D Berkeley Striping/Datamesh, Petal many Berkeley, Purdue (COMA) Linkabit, Hughes Qualcomm CERN, Illinois (Mosaic) Alta Vista Netscape, Yahoo, Google CMU, SRI, MIT Bell, IBM, Dragon Dragon, IBM Stanford, UCLA Bellcore (Telcordia) Amati, Alcatel, Broadcom ~ University ~ Industry R&D · ~ ~ - - - Products 39 bit. market |
40 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE from a research result to advanced development to product development to eco- nomic return. Ideas and people tend to bounce around; new ideas can be stymied by political or business impediments and forced to take alternative routes. The indus- try does not have a few mechanisms for creating and exploiting innovations. In- stead, it enjoys what has been called a "national research culture" that fosters innovation (Lazowska, 1998~. Some of the features of that culture are de- scribed below. A 1995 report of the National Research Council Computer Science and Telecommunications Board documents the effect of the research culture in the computer and high-performance computing arenas (NRC, 1995~. A more recent report (NRC, 2003), documents how technologies are born (often in academia), are taken up and extended by other academic or industrial groups, become the seeds of start-up companies or new products in larger companies, as well as how the market for the technology grows and matures (Figure 2-1~. All paths to market are erratic, and often take 15 years. The diversity in the academic and industrial sectors lends robustness to the process: a good idea is very hard to completely eradicate. Among the findings of that report are: Research has kept paying off over a long period. The payoff from research takes time. At least 10 years, more often 15, elapse between initial research on a major new idea and commercial success. This is still true in spite of today's shorter product cycles. · Unexpected results are often the most important. Electronic mail and the "windows" interface are only two examples. Research stimulates communication and interaction. Ideas flow back and forth between research programs and development efforts and between academia and industry. Research trains people, who start companies or form a pool of trained personnel that existing companies can draw on to enter new mar- kets quickly. Doing research involves taking risks. Not all public research programs have succeeded or led to clear outcomes even after many years. But the record of accomplishments suggests that government investment in com- puting and communications research has been very productive. Mobility of Ideas As Figure 2-1 suggests, the number and types of research structures among universities and industry provide a kind of redundancy; an idea that cannot ad- vance in one environment may flourish in another. As an example, the path of reduced instruction set computing (RISC) started out with John Cocke's IBM 801, developed at IBM Research. Although the ideas were countercultural and
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 41 did not have a great impact at IBM, they spawned two research projects, one at Berkeley and one at Stanford, to explore them further. Both university projects resulted in prototype processor designs. The Stanford project formed the nucleus of a start-up company MIPS Computer to build RISC microprocessors. The Berkeley project led to an advanced development project at Sun Microsystems to develop its S PARC instruction set. Both led to commercially successful products. Moreover, publication of the work in professional journals rapidly spread aware- ness of the technology. A related example occurred in the evolution of relational databases. A researcher at IBM, Ted Codd, developed an idea that found little encouragement at IBM, whose products at the time used a competing database technology. Nevertheless, Codd was able to seed academic work at Berkeley that enlarged the interested community and eventually led to two start-up companies, Ingres and Oracle, and a huge industry (NRC, 1999~. Both processor and data- base technologies were later embraced by IBM. Table 2-2 shows one way that ideas can move between academia and industry. A key idea of this sketch is the "democratization" phase, in which a tiny research community is deliberately enlarged into a community with a critical mass of researchers exchanging ideas, building prototypes, and teach- ing others. This step was clearly discernible in the histories of both RISC and relational databases, and was, perhaps, the key step in the spread of very large- scale integration design techniques that Carver Mead and Lynn Conway devel- oped in the late 1970s. Democratization in that case involved writing a text- book, teaching teachers, and starting courses in a half-dozen graduate departments to spread the ideas. These efforts resulted in a self-sustaining research community that built computer-aided design software, built a short- run chip fabrication system (MOSIS), designed a number of novel chip archi- tectures, and trained hundreds of students in the art of integrated-circuit design and computer architecture. TABLE 2-2 How Ideas Can Move between Academia and Industry University Industry/Government 1. Theoretical result ~ 2. Student graduates to industry laboratory ___ ~~ that encourages individual researchers ___ ~ and builds a basic prototype. 3. Democratization phase, in which many people work on the idea (e.g., RISC) ~ 4. Advanced development leads to a commercially successful product in a __ small but significant market. 5. Academics study the details and fill in ~ 6. Market explodes. Industrial research the gaps [lots of difficult research here] advances technologies. Source: Tennenhouse, 1998.
42 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE Mobility of People People often move on to new challenges, sometimes taking with them inno- vative ideas. The industry asserts, "technology transfer is a contact sport," that ideas transfer best when people carry them. Universities are, of course, a primary source of people, graduate and undergraduate students, many of whom have had research experience. University graduates with research experience are very valu- able to industry, not only as staff for its research functions, but also as technical leaders in product development organizations. Curious people with broad techni- cal knowledge who are trained in solving technical problems through research are extremely valuable in today' s product engineering groups. People with ideas often feel impelled to find a receptive environment, and in the 1990s start-up companies were a powerful attraction. University graduates sometimes embarked immediately on a start-up company, based on ideas formu- lated or prototyped while they were students. Faculty members often took leave to start companies or to consult with companies that embraced their new ideas. In fact, departing faculty members and students left some academic computer sci- ence departments with large gaps in their curricula and research programs, espe- cially systems and networks programs (Morris, 1998~.3 The lures are not only financial. Many students who would otherwise go to graduate school complain that academic research is sterile and irrelevant; they prefer to actually do engi- neering, to build a product that will change the world. The mobility of people includes flow from industry to academia. For in- stance, in a 2001 survey, many computer science programs reported record num- bers of applicants for doctoral programs, attributable to the demise of many Internet start-ups (Bryant and Vardi, 2002~. Anecdotal evidence also suggests that many industry researchers have found places in universities as industrial research spending has declined at telecommunications and computer hardware companies. Their knowledge and industrial experience can provide valuable in- sight for academic research. Open Structures The structure of the network systems and communications industry promotes certain kinds of innovation. When components have well defined interfaces, innovators can offer improved components with the same interfaces, so they remain compatible with their predecessors. Thus, there is a ready-made market for the new product. Moreover, some interfaces are specifically designed to ac- commodate innovation. For example, the "operating system interface" invites application programmers to write innovative applications by eliminating the need to deal with a myriad of details of hardware control. Standards for communica- tions and network protocols (interfaces) allow innovative products to interoperate with or supplant predecessors.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 43 An "open" interface by definition is an interface controlled by a consensus of the interested parties. Formal standards, for example, are open, and the industry has many industry consortia that develop and maintain open standards (e.g., IETF, X/Open, OMG). Open interfaces promote innovation by providing innova- tors with a dependable, stable interface for new products. The success of the deregulation of telecommunications services depends on standard interfaces. For example, a compatible local exchange carrier must be able to connect into the networks of other local exchanges and long-distance carriers with predict- able interfaces. Software Software is the universal building material of network systems and communi- cations products and services, and the importance of software as an enabler of innovation cannot be overemphasized. The vital functions of many products and services are controlled by software. Many innovations are merely software im- provements, but dramatic innovations sometimes come from relatively simple soft- ware. For example, the World Wide Web is essentially a set of common standards applied to a preexisting Internet communications infrastructure enabled by browser software. Another example is an easy change in software to use a digital compres- sion algorithm to increase the effective speed of a modem; this change was based on digital signal processors becoming fast enough to implement a high-speed mo- dem in software alone. An encryption feature could also be added easily. Software coupled with telecommunications has another virtue updated soft- ware can be distributed rapidly to customers over a network. Whether the new version fixes a bug or introduces a new feature, customers can easily upgrade their equipment, and new software systems are frequently introduced by allowing free downloads over the network (e.g., the Netscape browser, media players). By distributing software widely at zero or low cost, firms count on network effects to generate even broader use and to build a strong market base for their products.4 Software can be customized to meet the special needs of individual cus- tomers. Although many vendors do not offer variants because of the expense of testing and maintaining separate versions for separate customers, "open source software," makes the source code available to customers allowing them to inno- vate independently. Although some large pieces of software (e.g., the Linux operating system and Gnu software tools) are available in open source, it is too soon to tell whether open source software will become a significant pattern in the industry. Multiplier Effect of Infrastructure lntrastructure is critical to the advancement of network systems and com- munications technologies. At any given time, the installed networking and
44 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE computing facilities are the substrate upon which further innovations are devel- oped and introduced. The innovations may, in turn, lead to pressure to enhance the infrastructure, thus initiating a new cycle. The early ARPAnet is an example; the need to connect ARPAnet to other networks led to internetworking proto- cols, most notably the transmission control protocol and Internet protocol (TCP/IP) (serf and Kahn, 1974~. The demand for connections led to higher transmission speeds, faster routers, and routing protocols that could be scaled to a larger network. The larger network and its protocols and naming conventions, in turn, provided the near-universal connectivity that led to the creation of the World Wide Web. The growth of the Web increased the demand on the capacity and scale of the network, and today, the infrastructure is being challenged to carry traffic with real-time requirements, such as VoIP (voice over IP) and video. Advances in network infrastructure have been a key to fostering innovation. The federal government made the initial investments in ARPAnet and NSFnet. As the network expanded to nonacademic customers, regional network consortia built up the network. Today, service revenues support the network, but the federal government continues to support experiments that may lead to significant im- provements in performance (Internet2~. A similar pattern of infrastructure invest- ment occurred in academic computing facilities. When workstations were first introduced, the National Science Foundation (NSF) helped equip academic re- search centers with the new technology, which served as a substrate for academic research in networking and interactive computing. Subsequently, parallel com- puters were provided to encourage research on software tools for writing high- performance parallel computing applications. Intellectual Property In addition to patents, the industry also issues many licenses and cross- licenses. No company has a dominant position in the industry based on intellec- tual property (in contrast to the way Xerox dominated the copier industry when its basic xerography patents were still in force). Patents covering interfaces- whether computer buses or communications protocols must be licensed widely because interfaces must be open to be widely used. Therefore, to receive their endorsement, most standards bodies require that patents covering standards be licensed liberally. Size of the Research Investment Government Funding The federal government has provided major support for university comput- ing and networking facilities. The Defense Advanced Research Projects Agency
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 45 (DARPA) and NSF were particularly active in the development and growth of the Internet and continue to provide the bulk of support for new initiatives in networking research and infrastructure. Federal funding for research in com- puter science increased from $129 million in 1980 to $1.5 billion in 1999 (NSF, 2001~. In 1999, roughly 33 percent of this funding was provided to universi- ties the rest went to industry and government agencies; more than 75 percent of funding for basic research went to universities. In electrical engineering, of which communications is a subset, federal funding for research remained basi- cally flat throughout the 1980s and 1990s, peaking at $881 million in 1993 and retreating to $699 million in 1999 (NSF, 2001~. The share of total funding for electrical engineering research that went to universities rose, however, from 10 percent to 27 percent during this period. The two federal agencies that support research in computing and communications are the U.S. Department of Defense (DOD) and NSF, in that order. Federal funds support roughly two-thirds of university research in computer science and electrical engineering. Some of this funding is used to support the acquisition of research equipment and to support graduate students. The number of graduate degrees in electrical engineering and computer science increased rapidly through the 1980s and the 1990s; the number of master's degrees awarded more than doubled; 925 doctoral degrees in computer science and 1,596 in elec- trical engineering were awarded in 1998 (Hill, 2001~. The proportion of nonresi- dent aliens in total doctoral degree enrollments in computer science and computer engineering has risen steadily since 1945, up to 55 percent in 2001. Interestingly, data from the most recent Taulbee Survey indicate that only 17 percent of new faculty are nonresident aliens; proportionately fewer foreign students take posi- tions at U.S. universities (Bryant and Vardi, 2002~. Industry Funding Computer-related industries tend to be R&D intensive. Firms in this sector spend a greater percentage of sales revenues on R&D than any other industry except medical devices and pharmaceuticals. In computer-related industries, roughly 10 to 20 percent of corporate R&D funds are spent on research (rather than development). According to a 1999 report by the National Research Coun- cil, "Such expenditures tend to derive from, and result in, the fast pace of innova- tion characteristic of the field" (NRC, 1999~. Although the volume of R&D investment in computer-related industries has kept pace with the growth of busi- ness over the past decade, the R&D spending of the telecommunications compo- nent of the network systems and communications sector has contracted in the wake of AT&T's divestiture, deregulation, and most recently, deep recession in the telecommunications industry. Although the amount of industry support for university research in network systems and communications is not known, overall industry support for research
46 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE in science and engineering in universities represents about 7 percent of the total universities receive. The percentage is higher, perhaps as high as 15 to 20 per- cent, in engineering; support varies by the rank and reputation of the university program (Morgan and Strickland, 2000~. For the most part, computer-related industries have tended to draw on academic research more extensively than the telecommunications industry. Some of the largest firms in information technology provide significant sup- port for university research. IBM currently spends several hundred million dol- lars per year for research at universities. Support for university research provides IBM with access to activities at universities and contact with potential future employees. Microsoft has established a large research organization that empha- sizes fundamental research. It too has established research collaboration with a number of universities, including Southern California, Utah, Yale, and West Virginia University. Other firms have established relationships with a small num- ber of universities. One example is the AT&T Center for Internet Research at UC-Berkeley that was funded by AT&T in 1998 for three years. Another ex- ample is Intel Corporation, which has established research sites at the University of California at Berkeley, Carnegie Mellon University, and the University of Washington among others (http://www.intel-research.net/~. Some research organizations require or encourage both industry and govern- ment funding. NSF supports university-based engineering research centers and science and technology centers that must have industry contributions to supple- ment government funding. Initiatives like Internet2 and Next Generation Internet, which are funded principally by government, solicit industry support. CONTRIBUTIONS OF ACADEMIC RESEARCH Academic research has made essential contributions to the network systems and communications industry. The special strengths of universities are reflected in the ways they have contributed to the industry: . . . Human capital. Undergraduates and graduate students educated in univer- sities have become key players in industry as individual researchers, devel- opment engineers, technical leaders, and entrepreneurs. Research experi- ence in universities is highly valued by industry even for nonresearch employees. As students and faculty flow to industry and start-up compa- nies, they provide an effective form of "technology transfer." Long-term fundamental research. With proper funding, academic re- search is able to work on long-term problems that may be ignored by industry or may even be anathema to dominant industry businesses, tech- nologies, regulations, or prejudices. Intellectual diversity. Academia provides an open setting that can en- gage colleagues in various disciplines and industries; the results are
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 47 reported in the open literature. Concurrent research projects and differ- ent approaches provide a kind of redundancy and expand the commu- nity of researchers on promising topics. Shared artifacts of experimental research, especially software, are an important way to disseminate re- search results. · Collaboration with industry. Direct collaboration between industry and academia, both on specific projects and in longer term relationships, has produced significant contributions to network systems and communica- tions. There are many collaborative structures but no dominant or "best" collaboration scheme. Test beds. University laboratories can serve as test beds for new tech- nologies. Most of the early participants in the ARPAnet, for example, were universities, which played an important role in testing and refining the technology. The pattern has continued with the Gigabit Testbed, vBNS, and other networks, such as a campus-wide wireless network at Carnegie Mellon University. · Nuclei for start-up companies. University research can lead to tech- nologies and people that become the seeds of new businesses. Examples are Google and Yahoo, both spin-offs of research at Stan- ford University. . At an October 1998 NAB workshop to collect information and exchange ideas for this study, the participants came to the following conclusions: (1) the network systems and communications sector has benefited greatly from a na- tional research culture in which individuals move frequently between academia and industry, thereby increasing their knowledge of both and their contributions to both; (2) personal relationships are crucial; and (3) universities not only in- vigorate the research culture with fresh students each year, but they also house oven research projects that anchor technical disciplines. Human Capital Industry looks to universities to educate and train students who will staff industry R&D projects. Industry considers human capital to be the most impor- tant product of universities even more important than new knowledge captured in research results. The question of whether industry wants students with a broad technical education or with training in specific skills, such as programming in a given computer language or the operation of a certain kind of computer or com- munication device, is answered differently by different businesses. Larger com- panies tend to prefer broadly educated candidates who can learn skills quickly on the job. Smaller companies that do not have people to serve as mentors and trainers prefer trained candidates.
48 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE Training in research is extremely important to innovation, even if an indi- vidual does not continue to do research. Industry considers research experience valuable because it demonstrates the abilities necessary for any technical en- deavor: self-motivation, problem solving, teamwork, knowledge of related re- search, contact with other researchers and colleagues, the ability to organize an amorphous problem, and the perseverance to overcome setbacks. Graduates with advanced degrees have already shown greater than average ability; and research training is considered evidence that an individual can tackle difficult technical problems, such as designing and building complex systems. Students of electrical engineering and computer science have typically been in great demand, not only by companies in the network systems and communications sector, but also by other companies trying to modernize their information technology. The Bureau of Labor Statistics estimates that almost 4 million information-technology workers will be needed by 2010, compared with 1.9 million in 2000 (Bureau of Labor Statistics, 2001~. If this projection remains accurate, the current rate of graduation in U.S. universities approxi- mately 27,600 undergraduates and 900 Ph.D.s in computer science per year- will not bridge the gap (Hill, 2001~. Students and faculty who participate in start-up companies are important to the culture of innovation. A significant number of network systems and communications businesses have been founded by students straight from universities or by faculty, who either take a leave of absence or leave permanently. These companies are often formed to exploit a technology developed in the university. One of the best-known examples is Sun Microsystems, which began as a start-up company to commercialize a com- puter workstation designed at Stanford and Unix software originally conceived and developed at Bell Laboratories by Thompson and Ritchie then further developed by Bill Joy at UC-Berkeley. In the summer of 1998, six (out of 60) members of the electrical engineering faculty were on leave from Stanford University to work with start-up companies. Faculty members who return to the university report that their research has been stimulated greatly by their experi- ence. A founder of Granite Systems, for example, said that he now has a far better sense of what it takes to produce a product, as well as the state of the industry (D. Cheriton, Stanford University, personal communication, Septem- ber 8, 1998~. A report by BankBoston on the impact of a research university concluded with the following statement: "If the companies founded by MIT graduates and faculty formed an independent nation, the revenues produced by the companies would make that nation the 24th largest economy in the world. The 4,000 MIT- related companies employ 1.1 million people and have annual world sales of $232 billion" (BankBoston,1997~. MIT-related firms are especially prominent in electronics and software. More than half of the companies founded by MIT graduates were founded by graduates in electrical engineering (which at MIT includes computer science).
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY Research in Engineering and Computer Science 49 University research in electrical engineering and computer science has made significant contributions to the network systems and communications industry. In some cases, academic research projects have been essential to the creation of billion-dollar businesses (see Figure 2-1 for some examples). Academic research has also built a foundation of techniques and analysis tools that are widely used as enabling tools by the industry. These include tech- niques for the optimization of computer programs, for the automatic design of integrated-circuit chips, and for the verification of bus specifications. These tech- niques are not dramatic developments that spawned businesses, but they have been important to the industry as a whole. The digital cryptographic techniques widely used today to ensure privacy and authentication in electronic commerce and related applications were invented in academia (NRC, 1996~. The develop- ment of object-oriented programming, from the first step (Simula 67) to the most recent form (Java), took 30 years. Most of the research was necessarily conducted in academia, because industry typically does not invest in risky research that offers only long-term prospects for payoff. The two case studies below illustrate how academic research has contributed to the network systems and communications sector: (1) the Internet, which shows a 30-year trajectory of academic and industrial R&D to build a revolutionary communications technology; and (2) research in signal processing that led to a start-up company, Amati Communications, that successfully exploited the technology. Case Study: The Internet Academic research played a key role in the development of internetworking, the connection of disparate networks into a worldwide, scalable, packet-switched network. The Internet, which now connects more than 100 million people and computers, was the direct result of government-funded research begun in the late 1960s to link different kinds of academic research computers. Although industry was essential to the scaling of the Internet and the development of services, the early technical development depended almost entirely on university research (for a more detailed case study, see SRI International, 1997; for a brief chronology, see Box 2-2~. The ARPA funded research, development, and deployment of this revolutionary packet-switching technology, because the telecommunications in- dustry showed no interest in participating. The story begins with the ARPAnet, which was initiated by DARPA in the late 1960s as a way to share access to expensive or special-purpose research computers around the country. Precursor ideas for packet-switching networks had been developed at MIT and UCLA, but Paul Baran is credited with discover- ing packet switching while at the RAND Corporation in the early 1960s. Donald Davies, a researcher at the U.K.'s National Physical Laboratory independently
50 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE discovered the idea of packet switching in 1965, deciding upon some of the same parameters for his network design as Baran, such as a packet size of 1024 bits. ARPA built an early network by contracting with Bolt, Beranek, and Newman to build packet switches (IMPs) at research computers at about a dozen universities. In addition, academic research projects were initiated to develop protocols by which different types of computers could communicate, to outfit the computers with suitable hardware and software interfaces to the network, and to measure the performance of the operating network. Similar networks were also developed, such as a network using satellite or radio-transmission links to connect the packet switches. Early protocol experiments, together with the clear need for interconnecting the various kinds of networks being developed, pointed to a need for "internetworking." The key idea is the Internet datagram, a universal way of formatting network packets, together with associated protocols (TCP/IP), intro- duced in a paper by Cerf and Kahn (1974) while Cerf was a member of the Stanford faculty. This paper provided the first definition of Internet architecture and led to implementations and experiments at several universities. With the TCP/IP implementation developed in "Berkeley Unix" software (at the Univer- sity of California at Berkeley) and released in the late 1970s, it was easy to connect academic research computers to the network. Ad hoc committees of academic researchers refined the TCP/IP protocol standards, including applica- tion protocols. In the late 1970s, academic computer science research centers not served by the ARPAnet banded together to form CSnet, also using the TCP/IP protocol. In
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 51 1983, the new network was linked to the ARPAnet, an event that could be called the birth of the Internet. Subsequently, the network continued to grow, and the demand for connections increased. In 1986, with the creation of NSFnet, the responsibility for the principal "backbone" of the nationwide network shifted from DARPA to NSF. The network could be used only for research and educa- tion, and academics continued to play a major role in network governance ~ · . and englneenng. With the emergence of the World Wide Web, the Internet was no longer only for research and education but became a worldwide network connecting busi- nesses and consumers, as well as researchers. The idea of browsing text docu- ments obtained in a uniform way from any machine connected to the Internet was developed in 1991 by Tim Berners-Lee, then at the CERN nuclear research facility in Geneva, Switzerland. Documents on the Web may contain "hyperlinks" to other documents, thus linking documents into a complex "web." Marc Andreesen and other researchers at the University of Illinois later extended the Web to include pictures and other types of data. They also built a graphically oriented browser, called Mosaic, that allows users to "click" to follow a hyperlink, thus opening browsing to a wide range of people. Jim Clark recruited Andreesen to cofound Netscape Communications, which developed Netscape's Navigator browser product based on Mosaic. Microsoft soon developed a competing prod- uct, Internet Explorer. The combination of pictures, ease of use, and supported products enabled the Web to grow with astonishing speed. It quickly developed into a mechanism for publishing, for finding information, and for transacting business electronically. As more computers were connected to the network, the bandwidth and switching capacity had to be expanded. DARPA and NSF, with university and industry support, organized a series of test beds to explore high-speed networking technologies and test emerging products and protocols. Between 1990 and 1994, NSF and DARPA funded the Gigabit Testbed Initiative, a university-industry- government effort to explore networking technologies at speeds of 155 Mb/s and higher; one of these test beds achieved long-haul transmission at 800 Mb/s. NSF operated the vBNS network (very high-speed backbone service) in conjunction with MCI to link more than 75 universities in a network with backbone speeds of 622 Mb/s to 2.4 Gb/s and access links of 43 Mb/s to 155 Mb/s until the vBNS network was terminated in April 2000. The participants explored new applica- tions of advanced communication bandwidth and protocols, as well as opera- tional and governance issues. Universities are presently engaged in a new round of infrastructure enhancement, Internet2, designed to meet a full range of aca- demic research needs. As the Internet expanded, commercial businesses and services grew up along- side government and academic operations. New companies were started to sell packet switches (routers), application software, authoring tools, and network services. The leaders were not the telecommunication companies, but start-up
52 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE companies, such as Cisco and Netscape, using technologies developed in univer- sities. Network service providers emerged as regional networks (e.g., NEARnet, BARRnet) and were encouraged to connect to the national backbone operated as part of NSFnet. International connections were also developed. As the Internet grew, major telecommunications companies began to offer Internet service as well. In the United States today, businesses and residents in most densely popu- lated areas have a choice of several Internet connection methods and ser- . . vice providers. The governance and engineering of the Internet are unique: a governing body (the Internet Activities Board) and engineering standards organization (Internet Engineering Task Force) consist of volunteers from Internet participants. At first, committees of academics and the few government contractors that were building and operating the network set protocol and other engineering standards. Today, these committees have broad participation from academia, industry, and non- profit organizations. Academic Contributions An SRI study commissioned to analyze the nature of the research that con- tributed to the Internet described the contribution of academic research in some detail (SRI International, 1997~: The Internet appears, overall, to be primarily a problem-driven, technology-based innovation that required little direct input from fundamental research for its real- ization. The driving forces, interestingly, were not profit incentives in the private market, but public goods, first in the realm of national defense and subsequently in the university and government research infrastructure, as a means of fostering communication among computer scientists. What we are calling the Internet's intrinsic technologies network design, packet switching, routers, protocols, browsers were the products of problem-driven research conducted in universi- ties and government contractor laboratories with government support. One pos- sible exception is the research conducted at the University of Illinois' NCSA [National Center for Supercomputer Applications], which took place in an envi- ronment (according to Andreesen) that enabled researchers to head off in direc- tions that looked "interesting" without seeking justification. Nonetheless, the con- text was one of application, as suggested by the Center's name. Although the evolution of the Internet did not encounter technical roadblocks that required fundamental research for their resolution before further progress could be made, there is obvious, fundamental research content in both the Internet's intrinsic and supporting technologies. The electronic and physical infrastructures that comprise the Internet clearly depend on information theory, solid-state physics, electro- optics, and other fields on which modern communications technology is based and for which NSF has provided substantial support. The SRI study stresses the importance of government, industry, and univer- sities in the development of the Internet and points out that, as the focus of
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 53 technical and organizational innovation shifted from government to industry over a 30-year period, universities played a constant supportive role. By following the career trajectories of some key individuals who moved among the three sectors, the SRI study highlights the importance of networks of individuals and the im- portance of human capital. The study also cites the frequent opportunities for interactions and linkages among the three sectors and the ease with which indi- viduals can move across sectors in the United States as important factors to success (SRI International, 1997~. The Internet is an example of the major impact academic research can have in the creation of an entirely new technology. Although the enormous impact of the Internet is unusual, this case illustrates how government, academia, and in- dustry can contribute to technological and therefore economic success. Case Study: Amati Communications Amati Communications was founded in 1991 by Stanford University profes- sor John M. Cioffi to commercialize technology for transmitting high-speed digi- tal signals over copper telephone wires. The technology, named discrete multitone (DMT) modulation, was one of several technologies competing to be adopted as an industry standard. DMT is now an accepted standard for providing DSL (digi- tal subscriber line) service and a commercial success. Research The original work on DMT was conducted in 1987 by a research group at Stanford directed by Cioffi, who was then an assistant professor. Cioffi used funds from an NSF Presidential Investigator Award (1987-1992), with matching funds from several companies, including Bell Communications Research, to in- vestigate asymmetric digital subscriber lines. The initial objective was to develop reliable transmission of high-quality digital movies over phone lines, which re- quired speeds about 10 times faster than integrated service digital network (ISDN) lines, the existing technology. Later, the objective evolved to encompass high- speed Internet access and other data applications. The researchers investigated many methods and focused on an old encoding technique called multitone trans- mission, in which separate frequency channels (tones) carry separate digital sig- nals. A crude analogy would be sending several channels of Morse code over a telephone line, with each channel using a different audible frequency (like the seven separate tones used in touch-tone dialing). A receiver can split out the separate tone frequencies and decode each Morse sequence. The goal of Cioffi's research was to transmit data as fast as possible, which would require using many separate frequency bands and sophisticated signaling techniques (not Morse code!) in each band. This objective led Cioffi and his team to seek fundamental improvements in digital signal processing algorithms
54 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE that could be applied to various channels. But the most important innovation was the adaptation of each band to the band's transmission characteristics. In effect, DMT measures the properties of signals transmitted over each band and then allocates data accordingly. A band that attenuates signals less than another band carries more data. A band that introduces less noise than another also carries more data. DMT also measures and compensates for the transmission character- istics of each pair of copper wires. This complex scheme is practical because of inexpensive digital signal-processing hardware. Some of the key innovations for DMT were patented while Cioffi's group worked to refine, promote, and publish the method in IEEE journals and at American National Standards Institute standards meetings. Three of the patents, which are assigned to Stanford, are still valuable and are licensed throughout the telecommunications and data communications industry. At least one patent, on pioneering artificial intelligence techniques used in the adaptation, is considered necessary to comply with any of the existing or emerging DMT standards. In addition to patents, Cioffi and his graduate students acquired valuable know-how that would benefit any company that attempted to use DMT. Amati Start-up Company In 1991, after an unsuccessful search for corporate partners, Cioffi founded Amati Communications Corporation with Stanford University and a Stanford alumnus, Mr. Kim Maxwell, who was the company's first CEO. Amati's vision was to "get DMT on every phone line in the world." To be successful, Amati had to make DMT an industry standard, which involved working with national and international standards organizations and competing with other technologies. On March 10, 1993, after competitive testing against several other technologies, including an alternative promoted by AT&T, DMT technology was selected unanimously as a U.S. standard. As part of Stanford's contribution to the company, Cioffi was given three years of leave (with 50 percent leave spread over two years to make up the third year). In exchange for an exclusive licensing and sublicensing privilege on the DMT patents, Stanford received stock and a promise of royalties. In 1997-1998, Stanford re- ceived $7.9 million when it liquidated its holdings in Amati; royalties totaled more than $8 million in 2001 (Stanford University Corporate Guide, 2001~. Between 1991 and 1998, Amati employed several of Cioffi's graduate stu- dents, at least four of whom were directly involved in the Stanford research and had considerable knowledge of DMT. Stanford's then dean of engineering, Jim Gibbons, was chairman of Amati's Board of Directors until 1998. Amati went public late in 1995 and was a growth leader on the NASDAQ for approximately one year. In March 1998, Texas Instruments acquired Amati and the DMT license for approximately $450 million in cash. At the end of 2001, 3.6 million residential DSL (digital subscriber lines) using DMT technology were
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 55 installed in the United States, which was then projected to grow to 13.5 million by 2005 (InternetWeek, 2001~. Academic Research in Economics, Social Sciences. Management, Design, and Policy Academic research in a variety of nonengineering disciplines has also contrib- uted to the network systems and communications sector. This research has focused on how computers and communications systems fit into larger socioeconomic sys- tems. Examples include how information technology systems increase business productivity; the effects of e-mail on people and organizations; how people use the Internet; the effect of the Internet on family structures; the effect of prices on communications services; and the value of communication services to consumers. Some of these studies have formulated and tested models to explain the behavior of people and systems in operation. For example, the GOMS model of performance grounded in cognitive psychology (Card et al., 1983) has been used to design interactive interfaces. John Anderson has built several successful com- puterized tutoring systems based on a detailed understanding of how people model specific subjects (e.g., geometry and algebra) and the errors in these models. Results of social psychology research on electronic communication have been used to improve training for new users, in effect teaching them about the social norms and effects that spring from the technology. In some cases, the models only described what had been observed, but in a few cases they were used to predict the behavior of future systems. For example, models can predict how well people will perform simple interactive computer tasks. Although models cannot predict whether one chat room or e-mail system will be more popular than another or the details of e-mail usage, general principles can be learned. For example, people are generally less inhibited when using an elec- tronic communication medium than in face-to-face interactions. Observations such as this can lead to a better understanding of network systems and communications, but they cannot be used as a guide to design. At a workshop held in connection with this study, the consensus was that these research topics should be given more attention, especially as the services provided by network systems and communica- tions become more important and affect a greater portion of society. Economics, Policy, and Regulation Because communication systems have historically been operated as regu- lated monopolies, researchers in economics and policy were able to study them extensively. The work of one academic, Alfred Kahn of Cornell University, was used as a basis for the deregulation of several industries, including trucking, airlines, and communications (Kahn, 1970, 1971~. Nevertheless, a great many questions about economics and policy remain to be answered (see Box 2-3~.
56 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 57 The Telecommunications Policy Research Conference is an annual forum for scholars engaged in research on policy-relevant telecommunications and information issues and public-sector and private-sector decision makers engaged in telecommuni- cations and information policy. The wide range of topics at the 2002 conference reflected the intense academic interest in telecommunications policy. Topics included: comparative telecommunications policies in the United States and abroad; broadband deployment and uptake; spectrum management; computer and Internet security; wire- less communications standards; mergers and acquisition; intellectual property; basic research in telecommunications; mass media; and numerous other topics. University researchers presented the majority of research results at these sessions, reflecting the active involvement of academic research. Business and Management Business schools have long been concerned with how information technol- ogy can be exploited for the benefit of businesses (see Box 2-4~. For example, research on decision-support systems not only developed techniques for collect- ing and presenting relevant business data to management, but also compared the quality of decisions made with different kinds of supporting technology. The rapid development of the Internet has opened up new avenues for study, such as supply-chain integration, which uses the network to connect a manufacturer's process-planning system to the corresponding systems of its suppliers to ensure a smooth flow of the component parts required to fulfill orders. Success will de- pend on solving problems related to information technology, network protocols, and control theory. Electronic commerce will certainly face new problems that must be addressed. Optimizing network design in network systems and communications busi- nesses is a similar problem to the transportation problems studied by operations research. In the early development of the ARPAnet, attention was focused on optimization of network design. Today, the emphasis is on techniques to expand networks to meet burgeoning demand. Psychology and Social Sciences Research by psychologists and social scientists has focused on how people use computer and communication systems, the effects of these systems on people, how people interact with each other, and how they work in organizations. These studies are retrospective, conducted after the technology has been deployed long enough for transient behaviors to abate. De Sola Pool's classic book, The Social Impact of the Telephone (1997), is a fine example. Other examples are Computers in Class- room Culture (Schofield, 1995), Connections (Sproull and Kiesler, 1991), The HomeNet Field Trial of Residential Internet Services (Kraut et al., 1996), and The Second Self (Turkle, 1984), a study of personal interactions with computers.
58 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE Research has also helped guide the design of computer and communication systems. The Psychology of Human-Computer Interaction, a classic work by Card, Moran, and Newell (1983), showed how studies in cognitive psychology could be used to estimate human performance when interacting with a computer. These and other performance studies have influenced the design of graphical user interfaces. Ethnographic studies of the behavior of boys and girls at play have been used to inform the designs of many products.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY Design Research 59 Several universities have developed broad multidisciplinary programs aimed at harnessing developments in information technology to human needs. A leader in this area, the MIT Media Laboratory, brings together individuals from a broad spectrum of disciplines, including the humanities and fine arts, to conduct re- search and application development. For example, the News in the Future Project explored innovative ways to present the news to people using electronic media by tailoring content, presentation, and structure to the needs of the viewer. In addi- tion to developing prototype applications, the laboratory often works on funda- mental technologies, such as video compression or image understanding. MECHANISMS FOR UNIVERSITY-INDUSTRY COOPERATION Collaborations between industry and academia raise some obvious questions about the kinds of organizations and mechanisms that work best. A questionnaire on the subject sent to 60 researchers for this study elicited a wide range of responses. For example, one respondent felt that "centers which promote close interactions between academic researchers and knowledgeable industrial spon- sors are probably a prerequisite for making progress." Another mentioned several collaborative arrangements: joint research programs, like MIT's Project Athena; experimental test beds and university centers, like the NSF-supported supercomputing centers; and consortia. Another respondent felt that "Centers have an indifferent record in communications . . . I doubt that such forms of collaboration will ever be a success." Still others felt that the structure of the organization didn't make much difference as long as the participants understood each other's value systems. NSF has several programs to create university-based, industry-university research centers and engineering research centers (ERCs), both of which require industrial participation. The ERCs, which are designed to integrate research and education, have generally received favorable reviews (Parker, 1997~. However, the Telecommunications Research Center at Columbia University, funded by NSF from 1985 through 1995, was the only ERC established in the network systems and communications area. The network systems and communications sector does not have an institu- tion comparable to the Semiconductor Research Corporation, an organization that provides industrial support for university research relevant to semiconduc- tors, based on a 10-year technology "road map" to help guide research funding decisions (Bailey et al., 1998~. Although the establishment of a consortium of network systems and communications businesses has been discussed, nothing has come of it so far. A consortium of computer storage peripheral companies, Na- tional Storage Industry Consortium, has been established to support academic research through focused programs like optical storage. In addition, some firms
60 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE have targeted their research support for a small number of universities; for ex- ample, in late 1998, AT&T and the International Computer Science Institute at UC-Berkeley announced formation of the AT&T Center for Internet Research (ACIR), a multimillion-dollar research center that AT&T agreed to fund for at least three years (AT&T and International Computer Science Institute, 1998~. Recently, Intel has sited research operations at UC-Berkeley, Carnegie Mellon University, and University of Washington, all centers where researchers are Intel employees and university professors are engaged as laboratory directors and technical leaders (Intel, 2003~. The Microelectronics Innovation and Computer Research Opportunities (MICRO) Program in the University of California system is an example of a state government effort to encourage university-industry cooperation. The MICRO Program was established in 1981 by the state of California to support innovative research in microelectronics technology, its applications in computer and infor- mation sciences, and its necessary antecedents in other physical science disci- plines. The program is a partnership between industry and the state in which the state supplements industry funds and waives overhead on university research funding. In 2001-2002, 96 companies contributed approximately $6 million in cash and equipment to fund 98 different projects (MICRO, 2002~. In some cases, MICRO support has led to increased federal funding, as well as long-term part- nerships between universities and industry. For instance, after an initial concept phase, the RAID (redundant arrays of inexpensive disks) project at UC-Berkeley received MICRO support, which led to the creation of an industrial consortium in 1988-1990. The federal government became a research sponsor in 1990. By 1996, RAID was a $10 billion industry. In general, university-industry collaborative arrangements in network sys- tems have received mixed reviews. No structure has emerged as the "best," nor has any scheme emerged that works robustly in different circumstances. It ap- pears that strong personal leadership and a collaborative spirit between an aca- demic researcher and his or her industrial counterpart are the elements essential to success. The problem, of course, is that a good collaborative project can founder if one key individual (e.g., the "champion" in the firm) is transferred or moves elsewhere. FINDINGS AND RECOMMENDATIONS Academic research has made essential contributions to the network systems and communications sector. Although these contributions trained researchers, new technologies, algorithms, and prototype systems; early operating experience; studies of social and economic effects cannot be quantified, they have undoubt- edly had a substantial impact. In this industry, the academic ivory tower has been heavily populated by entrepreneurial engineer-researchers.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY Findings 61 Finding 2-1. Academic research has played and will continue to play an important role in the research culture of the network systems and communications industry. University-industry collaborations are fostered by a vigorous research cul- ture, and academic research has been crucial to the technical evolution of the industry, especially in the development and deployment of the Internet. To be sure, the recent deep recession in the telecommunications sector, which has forced significant reductions in corporate R&D budgets and manpower, has further diminished the industrial research contributions of this important subsector of the network systems and communications industry a trend begun with the breakup of the Bell system and further deregulation. Given the historical reliance of the telecommunications sector on internal industrial research, changes may be needed. If trends in industrial research persist, academic research in telecommu- nications will have to be increased. However, for the most part, the research culture that supports the network systems and communications industry is func- tioning well and needs no major repairs. People are the key components of this research culture. Collaboration be- tween universities and industry often depend on sometimes fragile personal rela- tionships that can be threatened if an industry researcher is reassigned or an academic researcher goes off in a new direction. Students, faculty, postdoctoral students, researchers with long-term visions, and researchers who focus on ap- plied problems play different roles. Contrary to popular opinion, university- industry projects are not devoted exclusively to long-term basic research; teams of faculty and students often address pragmatic, applied problems in close coop- eration with industry. The flow of people from academia to industry and vice versa is essential to the well-being of the industry and to academic research. The university's role of fueling the research culture with trained students is unique, and training in re- search is extremely important for innovation, even if the researcher does not continue to perform research but becomes part of an academic-like cadre that pursues innovations in industry (such as the groups awarded the Association for Computing Machinery's Software System Award). Universities also have a very broad research culture, and network systems and communications systems have increasingly drawn on the wide range of tech- nologies and expertise available at research universities. Electrical engineering and computer science are, of course, central to the industry, but other areas, such as cognitive science, social science, economics, and business modeling, are be- coming increasingly important, especially as the importance of information technology-delivered services increases. Some research universities (e.g., UC- Berkeley, University of Michigan, Indiana University) have created information- centered schools that focus on the social, political, and organizational context of information.
62 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE To participate in the research culture, a company must have a capacity to absorb innovation, an industrial research laboratory, for example, that can absorb people and ideas from outside the firm and exploit them within the firm. Despite the trend in industry research toward applied problems, industry laboratories have so far retained this absorptive capacity. In the last 15 years, innovations in the industry have focused on completing the digital revolution in telecommunica- tions (e.g., new switching gear); new transmission protocols (e.g., SONET and Asynchronous Transfer Mode), and transmission equipment; faster modems; the refinement and deployment of fiber optics; the refinement of IF protocols; and the widespread deployment of the Internet. The focus is now shifting toward innovative services, which requires an understanding of psychology, consumer behavior, social phenomena, and other disciplines that can inform the design and operation of new services. The explo- sive growth of the World Wide Web can be attributed to its social properties more than to its technical capabilities. Industry is most likely to devise and launch new services, but formal research on the uses and effects of these services is most likely to be undertaken in academia. Finding 2-2. Innovation cycles in the network systems and communications industry have worked well. An astonishing number of incremental changes have cumulatively taken on the character of breakthroughs. The Internet, for example, had its roots in the ARPAnet in 1969, was developed and deployed incrementally, and was launched into the public arena by the World Wide Web and browsers in 1993. The effect on the industry was of a breakthrough; the telecommunications industry today is utterly different from the industry of 10 years ago. Similarly, the incremental evolution of technologies (e.g., batteries, low-power circuits, integrated radio-frequency elements) to support small portable devices spawned breakthrough products, such as pagers, cellular telephones, and packet- radio modems. Innovations in the network systems and communications industry have been characterized by the integration of a wide range of technologies: chip designs with increasing levels of integration; digital-analog integration in wireless and wireline communications (e.g., cellular and satellite telephony, wireless devices, such as pagers and security devices, modems and cable modems, local-area net- work and intranetwork systems and communications receivers, optical network interfaces); internetworking technologies; and, above all, the increasing use of software technologies of all kinds. The convergence of computing and telecom- munications has brought together a wide array of technologies for new products. Academic contributions to these innovations have varied widely. Some inno- vations originated in university research and were spun off into venture-capital- backed start-up companies. This route is supported by established university policies, a workforce eager to engage in risky start-ups, and a mature venture-
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 63 capital industry that has been willing to back telecommunications and network- ing businesses. Many academics have been consultants for network systems and communications companies. Academic design departments have worked on in- dustry projects, and some companies have supported academic research, often directed toward solving specific problems. Sometimes, new companies or prod- ucts have emerged from business-school entrepreneurship programs. Finding 2-3. The success of industry-academic collaboration (as defined by participants) depends more on leadership and people than on the type of collabo- rative structure. Organizational structures, such as research centers that foster university- industry collaboration, receive mixed reviews in network systems and communi- cations. Success appears to depend less on the choice of structure, the funding arrangements, or the legal agreements than on the leadership and passion of the people involved. A committed leader is essential for establishing personal rela- tionships, and, in general, researchers consider inducements to individuals (as opposed to institutions) more effective than collaborative organizational struc- tures. Dependence on personal relationships can sometimes lead to problems, however. A collaborative effort between industry and a university can founder if a key individual (the "champion" within the firm) is transferred or moved else- where within the company. Overall, therefore, the diversity of approaches to industry-academic collaboration is healthy for both partners. Finding 2-4. Creating standards is an important aspect of innovation in commu- nication systems. Standards are necessary for interoperability, which is essential to the indus- try. The more interoperability, the faster the growth of the user base and the faster the increase in value of the system. The success of many businesses depends on the number of other entities that can communicate in a network the value of network externalities. Committees of industry members, sometimes with academic participation, usually determine standards. Many Internet standards groups, such as the Internet Engineering Task Force (IETF); the ATM Forum, which was organized to pro- mote data-networking uses of Asynchronous Transfer Mode (ATM); and the discrete multilane modulation standard for asymmetric digital subscriber lines, have academic participants. In the United States, university researchers often have difficulty participating in setting standards because of time and travel de- mands. In Europe, academic participation has been stronger. U.S. researchers could be helpful in gathering data and analyzing standards proposals; good data and independent analysis could reduce squabbling over business biases and focus more attention on design issues.
64 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE Finding 2-5. Academic institutions have been at the forefront of network infra- structure deployment. The United States led the way in deploying advanced infrastructure critical to enabling research (e.g., ARPAnet, NSFnet, vBNS, NGI, Internet2~. The de- ployed infrastructure has led to further developments. For example, the World Wide Web was successful partly because networking infrastructure had already been installed. Academic institutions have played a crucial role in the deployment of net- work infrastructure. Ever since NSFnet was formed, universities have recognized the importance of the Internet to academic endeavors of all kinds, not just com- puter science and engineering research. With funding from NSF and other sources, universities have been willing to deploy leading-edge technologies. Deployments such as Internet2, which was spurred in large part by universities, are likely to increase the speed and throughput of network services available to universities and thus to support research that requires high-performance networking infra- structure. However, this may not necessarily stimulate research on networking. Conflicting demands on these systems has created some tension between provid- ing robust services for other research fields and experimentation for networking research. A state-of-the-art network that can be used for experiments in network- ing has not been developed. Finding 2-6. The network systems and communications industry is evolving in directions that may require new kinds of university-industry partnerships to ex- ploit research. AS communicating appliances proliferate, the need for harmonizing the tech- nological and human elements increases. Examples include: designing communi- cation services that users can understand and exploit; integrating multiple devices and services to create personalized configurations; designing new user interfaces; and streamlining or automating customer service functions. Even within a net- work, areas that are not purely technical will also require research: the provision of services; the quality of service; incompatibilities between heterogeneous prod- ucts and services; security; and network management and operation. Optimal interactions among product design, network organization and management, ser- vice provision, and technology will require close collaboration between univer- sity researchers in the social, behavioral, and management sciences on one hand and engineers and scientists on the other. As the industry moves toward offering more "information services" rather than "communications devices," it must turn to the market rather than to research for guidance. In the future, university re- search might focus on how individuals and society as a whole value and use network systems and communications services. Finding 2-7. Many Internet service providers are not willing to make their data available to researchers.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 65 Many pressing research questions in network systems and communications require studies of the characteristics of networks under normal operating condi- tions as a basis of comparison. Research on limiting congestion, improving the quality of service, and improving routing requires traces or other logs of actual network activity. Although data are available for some experimental networks, many Internet service providers have not been willing to make their data avail- able to researchers. This has hampered university research that might lead to improved network design and operation. The problem could get worse if service providers become more vertically organized and less open about their operations, problems, and needs. Finding 2-8. The high cost of protecting intellectual property could im- pede research. Industry and academia are both becoming increasingly protective of their intellectual property rights, and the enormous economic activity in the network systems and communications sector encourages this trend. The processes of working out licensing and sharing agreements could impede the free flow of ideas necessary for research to flourish. Whenever universities band together in a research consortium or a single university-industry collaboration is started, researchers spend much time and energy working out intellectual prop- erty agreements. The trend can be counteracted in several ways. "Open source" software licenses that implicitly recognize that unused intellectual property has no value explicitly promote sharing. Other ways of reducing the costs of intellectual property protection could include standardized forms of collaborative research agreements. Finding 2-9. Long-term research is important to the future of the network sys- tems and communications industry. Despite the apparent success of network systems and communications tech- nologies, many difficult problems must be solved for the industry to continue to grow and prosper. Continued rapid expansion of a sophisticated communications infrastructure with millions or billions of network elements is bound to face some difficult problems. The industry needs software-engineering discipline to ensure that modules intended to fit together do so and that upgraded modules can be introduced without disrupting network operation. Distributed systems must be designed to be robust under failure, to remain stable under all operating condi- tions, and to guarantee performance requirements. Some long-term problems, such as the quality of service and security, are related to scale limitations. Current data networks require many people with sophisticated skills; and the design complexity and deployment scales of these systems exceed our engineering knowledge. As the extraordinary growth of the industry slows, new problems will have to be addressed, such as the impact of microelectronic components with different costs and technical properties, new computational needs as the rate of
66 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE miniaturization slows, and the technical implications of these changes. Indeed, as industrial research investments change, and as human capital stresses wax and wane, it is important to keep long-term academic research activities alive, pre- cisely because they are the long-lived seeds from which both ideas and people can spring, regardless of the short-term financial health of the industry. Recommendations Recommendation 2-1. Universities and industry should take steps to ensure that faculty and students are available to carry on research in computer science and other information technology fields in the future. Innovation, either from research or incremental engineering, depends on trained researchers. Projected demand for computer science and other information technol- ogy graduates indicates periodic shortages in coming years. To maintain the pipeline of both academic and industrial researchers, the following measures could be taken: . Universities should Provide early research experiences for undergradu- ates or even secondary school students. Universities should provide career-development support for young fac- ulty members. · Fellowships should be provided for graduate students to encourage them to pursue research degrees; industry should provide some of this support. · Universities and industry should provide incentives for industry engineers to return to academia for training in research. · Universities should develop cooperative programs in which master's de- grees are based not only on course work, but also on research experience. · Training in academic research should include training in some of the qualities students will need for jobs in industry. Research should involve addressing not only small technical puzzles in isolation, but also complex systems problems in context. Students should be encouraged to confront complexity and to address real-world data and operational problems. · Research should encourage teamwork. · High-caliber industry researchers and engineers should be encouraged to take sabbaticals to work in academia, thus bringing real-world research problems into academic settings. Recommendation 2-2. Universities and industry should continue to develop diverse collaborative arrangements.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 67 Industry and universities should resist the temptation to impose standard struc- tural mechanisms to promote collaboration. Incentives for personal interactions between university and industry should be encouraged in the following ways: · Provide support for strong, committed leaders and the collaborative orga- nizations they lead. Encourage sabbaticals in both directions, enabling academics to spend time in industry, especially in start-up companies. Support people and projects that involve academic and industry research- ers in essential ways. · Explore new ways to support personal interactions across academic- industry boundaries, including using technology to support collaboration. Recommendation 2-3. Universities and industry should make every effort to invigorate academic research on networking. The extraordinary success of the Internet and the lure of Internet-related start-up companies have tended to focus attention on short-term goals, caus- ing long-term research to suffer. The situation could be improved in the following ways: Acknowledge that the research community must take risks. Focus academic research on the thorny problems of large systems: model- ing, maintenance, upgrades, quality-of-service, security, and so on. Both funding agencies and academics must recognize that large-scale systems can best be addressed in a university setting. Even applied systems research can be structured in a way that accommodates a long- term approach. Universities and funding agencies (and industry) should support long- term, radical research on networks. Universities and industry should encourage interdisciplinary research that combines network technologies with design and social science disciplines. Networked devices (especially hand-held mobile devices) will have to meet both technical and human requirements. Universities should recognize that valuable innovations and engineering in the field are often not channeled through traditional peer-reviewed publications. Therefore, effective industry interaction should be more highly valued in decisions about academic promotion and tenure. To revitalize academic research on networking, the National Science Foundation should consider sponsoring a workshop on the subject that brings together academic and industry participants. A new agenda could provide a strong argument for industry support, either by individual firms or by a consortium.
68 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE NOTES 1 Gordon Moore (cofounder of Intel) predicted in 1965 that the transistor density of semicon- ductor chips would double roughly every year. See Moore, 1965. 2For example, during fiscal year 2001, Microsoft spent $4.38 billion on product research and development activities excluding funding of joint venture activity. This represented 17.3 percent of revenue that year. Microsoft Research, the part of the company that looks more than one or two product cycles out, has around 600 employees and a budget of roughly $200 million, less than 5 percent of the $4.38 billion, or less than 0.8 percent of total revenue. 3The loss of faculty to commercial endeavors was limited in time and to only a few programs. Data from the most recent Taulbee Survey of computer science and computer engineering depart- ments indicate that faculty numbers have grown and are anticipated to grow through 2004. The survey also indicates that faculty departures have ranged from 2.3 to 2.6 percent over the last several years (Bryant and Vardi, 2002). 4Economists have long acknowledged "externalities," factors that alter the value of a good viewed in isolation. Shapiro and Varian (1998) applied the idea to networks, so-called "network effects." Robert Metcalfe, a popular speaker on the value of networks, has often said that the usefulness, or utility, of a network equals the square of the number of users. This observation has been dubbed "Metcalfe's law" (Gilder, 1993). REFERENCES AT&T and the International Computer Science Institute. 1998. AT&T Labs, ICSI establish Internet research center. Press release October 8, 1998. Available online at: http://www.icir.org/ aciri.html. [June 24, 2003] Bailey, D.E., F.S. Settles, and D. Sanrow. 1998. Designing, Controlling, and Improving SRC Re- search Quality. Presentation at the NAE-Committee on Science, Engineering, and Public Policy Workshop on the Role of Human Capital in Capitalizing on Research, Beckman Center, Irvine, California, January 21, 1998. BankBoston. 1997. MIT: The Impact of Innovation. Special Report of the BankBoston Economics Department. Available online at: http://web.mit.edu/newsoffice/founders. [June 24, 2003] Bell, D.G., D.G. Bobrow, O. Raiman, and M.H. Shirley. 1997. Dynamic Documents and Situated Processes: Building on Local Knowledge in Field Service. Pp. 261-276 in Information and Process Integration in Enterprises: Rethinking Documents, T. Wakayama, S. Kannapan, C.M. Khoong, S. Navathe, and J. Yates, eds. Norwell, Mass.: Kluwer Academic Publishers. Bryant, R.E., and M.Y. Vardi. 2002. 2000-2001 Taulbee Survey: Hope for More Balance in Supply and Demand. Computing Research News 14(2): 4-11. Brynjolfson, E. 1991. Information Technology and the "Productivity Paradox": What We Know and What We Don't Know. Cambridge, Mass.: MIT Sloan School of Management. Bureau of Labor Statistics. 2001. Table 3b. Fastest Growing Occupations, 2000-10. Available online at: http://www.bls.gov/news.release/ecopro.tO6.htm. [June 24, 2003] Card, S.K., T.P. Moran, and A. Newell. 1983. The Psychology of Human-Computer Interaction. Hillsdale, N.J.: L. Erlbaum Assoc. Cerf, V., and R. Kahn. 1974. A protocol for packet network intercommunication. IEEE Transactions on Communications 22(5): 637-642. Cohen, W.M., and D.A. Levinthal. 1990. Absorptive capacity: a new perspective on learning and innovation. Administrative Science Quarterly 35(1): 128-152. CTIA (Cellular Telecommunications and Internet Association). 2003. CTIA's Semi-Annual Wire- less Industry Survey. Available online at: http://www.wow-com.com/pdf/CTIA_Survey_ Yearend_2002.pdf. [June 24, 2003]
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY 69 CTIS (Corporate Technology Information Services, Inc.). 1999. Spotlight. Woburn, Mass.: Corpo- rate Technology Information Services, Inc. Cusumano, M. 1991. Japan's Software Factories: A Challenge to U.S. Management. Oxford, U.K.: Oxford University Press. de Sola Pool, I., ed. 1977. The Social Impact of the Telephone. Cambridge, Mass.: MIT Press. Earl, M., ed. 1988. Information Management: The Strategic Dimension. Oxford, U.K.: Clarendon Press. Gilder, G. 1993. Metcalfe's Law and Legacy. Forbes ASAP, September 13. Available online at: http://www.seas. upenn. edu/~gajl/metgg.html. [June 24, 2003] Hill, S. 2001. Science and Engineering Degrees: 1966-1998. Arlington, Va.: Division of Science Resources Studies, National Science Foundation. Intel. 2003. Intel Research Network of Laboratories. Available online at: http://www.intel-research.net. [June 24, 2003] InternetWeek. 2001. Report: DSL Market to Rebound Next Year. October 24, 2001. Available online at: http://www. internetweek.com/story/INW20011024S0005. [June 24, 2003] Kahn, A. 1970. The Economics of Regulation, Vol. 1. New York: John Wiley & Sons. Kahn, A. 1971. The Economics of Regulation, Vol. 2. New York: John Wiley & Sons. Keen, P.G.W., and M.S. Scott Morton. 1978. Decision Support Systems: An Organizational Perspec- tive. Reading, Mass.: Addison-Wesley. Kraut, R., W. Scherlis, T. Mukhopa&yay, J. Manning, and S. Kiesler. 1996. The HomeNet field trial of residential Internet services. Communications of the ACM 39(12): 55-63. Lazowska, E. 1998. Remarks made during panel session on Changing the Interaction Between Aca- demic Research and Industry: University, Industry, and Government Perspectives. Presented at the workshop How Can Academic Research Best Contribute to Network Systems and Commu- nications?, National Academy of Engineering, Washington, D.C., October 30, 1998. Malone, T. 1998. Remarks made during panel session on Contributions and Impact of Academic Research: Social, Management, and Policy Sciences. Presented at the workshop How Can Academic Research Best Contribute to Network Systems and Communications?, National Academy of Engineering, Washington, D.C., October 30, 1998. Malone, T.W., K. Crowston, J. Lee, B. Pentland, C. Dellarocas, G. Wyner, J. Quimby, C.S. Osborn, A. Bernstein, G. Herman, M. Klein, and E. O'Donnell. 1999. Tools for inventing organizations: toward a handbook of organizational processes. Management Science 45(3): 425-443. Markus, M.L. 1987. Toward a "critical mass" theory of interactive media: universal access, inter- dependence, and diffusion. Communications Research 14: 491-511. McKnight, L.W., and J.P. Bailey, eds. 1997. Internet Economics. Cambridge, Mass.: MIT Press. Messerschmitt, D.G. 1996. The convergence of telecommunications and computing: what are the implications today? IEEE Proceedings (August). Available online at: http://www.informatik. tu-darmstadt.de/VS/Lehre/VES97-98/JavaTK/convergence.html. [June 24, 2003] MICRO. 2002. University of California Microelectronics Innovation and Computer Research Oppor- tunities. Available online at: http://www.ucop.edu/research/micro/. [June 24, 2003] Moore, G.E. 1965. Cramming more components onto integrated circuits. Electronics 38(8): 114- 117. Available online at: http://www.cybraryn.com/news/documents/moore-paper.pdf. [June 24, 2003] Morgan, R.P., and D.E. Strickland. 2000. U.S. university research contributions to industry: findings and conjectures. Science and Public Policy 28(2): 113-121. Morris, J. 1998. Remarks made during panel session on Changing the Interaction Between Academic Research and Industry: University, Industry and Government Perspectives. Presented at the workshop How Can Academic Research Best Contribute to Network Systems and Communica- tions?, National Academy of Engineering, Washington, D.C., October 30, 1998
70 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE NRC (National Research Council). 1995. Evolving the High-Performance Computing and Commu- nications Initiative to Support the Nation's Information Infrastructure. Washington, D.C.: Na- tional Academy Press. NRC. 1996. Cryptography's Role in Securing the Information Society. Washington, D.C.: National Academy Press. NRC. 1999. Funding a Revolution: Government Support for Computing Research. Washington, D.C.: National Academy Press. NRC. 2000. Making IT Better: Expanding Information Technology Research to Meet Society's Needs. Washington, D.C.: National Academy Press. NRC. 2003. Innovation and Information Technology. Washington, D.C.: National Academies Press. NSF (National Science Foundation). 2001. Survey of Federal Funds for Research and Development: Fiscal Years 1999, 2000, and 2001. Arlington, Va.: National Science Foundation. Orr, J. 1990. Sharing Knowledge, Celebrating Identity: War Stories and Community Memory in a Service Culture. Pp. 169-189 in Collective Remembering: Memory in Society, D.S. Middleton and D. Edwards, eds. Beverly Hills, Calif.: Sage Publications. Parker, L. 1997. The Engineering Research Centers Program: An Assessment of Benefits and Out- comes. Arlington, Va.: National Science Foundation. Qualcomm Corporation. 1999. ERICSSON and QUALCOMM Reach Global CDMA Resolution. Press release. March 25, 1999. Available online at: http://www.qualcomm.com/press/pr/ releasesl999/press457.html. [June 24, 2003] Rockart, J.F. 1981. The changing role of the information systems executive: a critical success factors perspective. Sloan Management Review 22(2): 15-25. Roessner, D., R. Carr, I. Feller, M. McGeary, and N. Newman. 1998. The Role of NSF's Support of Engineering in Enabling Technological Innovation: Final Report-Phase II. Arlington, Va.: SRI International. Available online at: http://www.sri.com/policy/stp/techin2/. [June 24, 2003] Schofield, J.W. 1995. Computers and Classroom Culture. New York: Cambridge University Press. Shapiro, C., and H. Varian. 1998. Information Rules: A Strategic Guide to the Network Economy. Cambridge, Mass.: Harvard Business School Press. Siegel, J.L., V. Dubrovsky, S. Kiesler, and T. McGuire. 1986. Group Processes in Computer- Mediated Communication. Organizational Behavior and Human Decision Processes 37: 157-187. Sirbu, M. 1998. Remarks made during panel session on Contributions and Impact of Academic Research: Social, Management, and Policy Sciences. Presented at the workshop How Can Academic Research Best Contribute to Network Systems and Communications?, National Academy of Engineering, Washington, D.C., October 30, 1998. Sproull, L., and S. Kiesler. 1991. Connections: New Ways of Working in the Networked Organiza- tion. Cambridge, Mass.: MIT Press. SRI International. 1997. The Role of NSF's Support of Engineering in Enabling Technological Innovation. IV. The Internet. Available online at: http://www.sri.com/policy/stp/techin/ interl.html. [June 24, 2003] Stanford University Corporate Guide. 2001. Top 10 Stanford Inventions. Available online at: http:// corporate.stanford.edu/innovations/invent.html. [June 24, 2003] Tennenhouse, D. 1998. Diagram presented during panel session on Changing the Interaction Be- tween Academic Research and Industry: University, Industry and Government Perspectives at the workshop How Can Academic Research Best Contribute to Network Systems and Commu- nications?, National Academy of Engineering, Washington, D.C., October 30, 1998. Turkle, S. 1984. The Second Self: Computers and the Human Spirit. New York: Simon & Schuster. U.S. Bureau of the Census. 2002. Statistical Abstract of the United States-2002. Washington, D.C.: U.S. Government Printing Office. Available online at: http://www.census.gov/prod/www/ statistical-abstract-02.html. [June 24, 2003]
71 ADDENDUM E-Mai} Questionnaire The following questionnaire was sent to individuals selected from various parts of the network systems and communication industry, some of whom at- tended the October 1998 workshop. Included among the questionnaire respon- dents were senior executives at AT&T Laboratories, Bell Atlantic, Bellcore, MCI, and Motorola, and professors with expertise in computer science and engi- neering, network systems, and telecommunications from Stanford University, University of Delaware, University of California-Berkeley, University of Cali- fornia, Los Angeles, University of Virginia, and University of Washington. THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE NETWORK SYSTEMS AND COMMUNICATIONS PANEL We invite your responses to the following questions. Your responses will be used by our Panel as background information for our report. Any material used verbatim will not be attributed to you without seeking your permission. 1. Could you describe briefly significant academic research contributions to the network systems and communications industry? (If possible, please supply references to published information that outlines the contributions.) 2. Overall, would you describe the impact of academic research on industrial performance in the network systems and communications industry as (Please put an X in one box): 1. very large 2. large 3. medium 4. small ~ 5. very small/non-existent 3. What is the role of academic research in educating people who work in your industry? (Please focus on university research activities, rather than univer- sity education generally.) 4. What structural forms of university-industry collaboration lead to good results in your industry? An example of such a structure might be a discipline- or industry-oriented "center" that solicits industry sponsors for a collection of
72 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE projects that span a varied research program. What seem to be the essential determinants of success of such structures? 5. What are significant emerging trends or problems that the network sys- tems and communications industry will face in the future that could benefit from academic research? 6. What changes are required, if any, in academic research if it is to be responsive to these industrial trends and problems? 7. What single step could be taken by universities to enhance the impact of academic research on the industry? 8. What single step could be taken by companies to enhance the impact of academic research on industry? 9. What single step could be taken by government to enhance the impact of academic research on industry? 10. Do you see any downside to enhanced university-industry research col- laboration? Things to be avoided? 11. Other comments? Any comments, pointers to other studies, or sugges- tions would be appreciated.
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY WORKSHOP AGENDA HOW CAN ACADEMIC RESEARCH BEST CONTRIBUTE TO NETWORK SYSTEMS AND COMMUNICATIONS? October 30, 1998 National Academies Building 2101 Constitution Avenue N.W. Washington, D.C. 9:00 am Welcoming remarks and self-introductions Wm. A. Wulf; President, National Academy of Engineering 73 9:15 am Overview of the work of the Network Systems and Communications Panel and description of the wider NAE study Bob Sproull, Panel Chair 10:00 am Break 10:15 am Session I. Contributions and impacts of academic research on performance in the network systems and communications indus- try: Engineering and the Physical Sciences David Forney, Ambuj Goyal, Robert Kahn, H. T. Kung, David Mills 11:45 am Lunch in Meeting Room 12:30 pm Session II. Contributions and impacts of academic research on performance in the network systems and communications indus- try: Design, Social, Management, and Policy Sciences Dan Atkins, Walter Bender, Robert Kraut, Tom Malone, Marvin Sirbu 1:30 pm Session III. Structures for university-industry collaboration James Flanagan, Stewart Personick, David Roessner, Donald Strickland, Stephen Wolff 2:30 pm Break
74 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE 2:45 pm Session IV. Changing the interaction between academic research and industry: University, Industry, and Government Perspectives Hamid Ahmadi, Ed Lazowska, James Morris, Rick Rashid, George Strawn, David Tennenhouse 4:30 pm Discussion, conclusions and recommendations Bob Sproull
NETWORK SYSTEMS AND COMMUNICATIONS INDUSTRY WORKSHOP ATTENDEES Robert Sproull, chair * Vice President and Sun Fellow Sun Microsystems, Inc. Hamid Ahmadi AT&T Labs Alfred V. Aho * Associate Research Vice President Communications Sciences Research Division Lucent Technologies Bell Labs Innovations Daniel Atkins School of Information University of Michigan Walter Bender MIT Media Lab John Cioffi * Associate Professor Department of Electrical Engineering Stanford University David J. Farber * Alfred Fitter Moore Professor of Telecommunications University of Pennsylvania James Flanagan Center for Computer Aids Rutgers University G. David Forney, Jr. Motorola, Inc. *Parley member 75 Ambuj Goyal IBM Corporation T.J. Watson Research Center George H. Heilmeier * Chairman Emeritus Bellcore Robert Kahn Corporation for National Research Initiatives Robert Kraut Department of Social and Decision Sciences Carnegie Mellon University H.T. Kung Department of Electrical Engineering and Computer Science Harvard University Ed Lazowska Department of Computer Science and Engineering University of Washington Tom Malone Sloan School of Management Massachusetts Institute of Technology David Mills Department of Electrical and Computer Engineering University of Delaware
76 THE IMPACT OF ACADEMIC RESEARCH ON INDUSTRIAL PERFORMANCE James Morris School of Computer Science Carnegie Mellon University Stewart Personick Drexel University Richard Rashid Advanced Technical Research Microsoft J. David Roessner School of Public Policy Georgia Institute of Technology Jerrard Sheehan National Research Council Marvin Sirbu Information Networking Institute Carnegie Mellon University George Strawn National Science Foundation Donald E. Strickland Chair, Management Department Southern Illinois University David Tennenhouse Defense Advanced Research Projects Agency Stephen Wolff * Executive Director Advanced Internet Initiatives Division Cisco Systems, Inc. Wm. A. Wulf President National Academy of Engineering NAE Program Office Staff Tom Weimer, Director Proctor Reid, Associate Director Nathan Kahl, Project Assistant Robert Morgan, NAB Fellow and Senior Analyst *Parley member