The NSFNET Backbone Project, 1987 - 1995
NSFNET: A Partnership for High-Speed Networking
The NSFNET Phenomenon
While the 1986 NSFNET had given the research and education community a taste of what was possible with connectivity to a high-speed backbone network, the leaps in capacity and reliability offered by the new NSFNET backbone service proved to be popular beyond even the partners' wildest dreams. The growth of the Internet worldwide, of which the construction of the NSFNET in the US played a major part, can only be described as "phenomenal."
"It was pretty much an unmanageable child," says Ellen Hoffman. "I think back then we all had hopes that it would grow the ways that it did, but none of us believed that it would grow so fast." Hans-Werner recalls working in an environment where "something that was brand-new and totally up-to-date in the morning was completely out of date by the afternoon. That happened pretty much on a day-by-day basis, because we moved so fast."
The statistics Merit provided on use of the NSFNET backbone are by now well-known, but still every bit as astounding: 19 billion packets per month processed in September of 1992, a growth rate of approximately 11% a month. Over 6,000 networks were connected to the NSFNET backbone service in 1992, one-third of which were outside the United States. Use of the network was also evolving; the share of network traffic devoted to electronic mail and file transfers was being overtaken by newer information storage and retrieval applications such as Gopher, Archie, and Veronica.
Support for the users of the network, from regional and midlevel networks to campus networks, was just as critical as the physical infrastructure, and Merit's Information Services group was tasked with giving the research and education community the background and tools they needed to use the network to its fullest. The idea of "user support" had been around since the "olden days" of campus computing systems, when "you have all this information--how do you get it so it is available to people? How do you help them use it, help them to find it?" explains Jane Caviness. Administrative information related to organization, funding and staffing of networks was also needed. Kathleen McClatchey, who as Director of Planning, Marketing and Public Relations at the University of Michigan, helped to lead the partners in these efforts from 1989 to 1991, explains that the task before them was to "try to take something at the 'bits and bytes' stage and focus on what it could actually do." Elise Gerich puts the question this way: "What makes the NSFNET real to people?"
The Merit Information Services staff met the challenge by conceiving, designing, and providing information services for a new community of users, whose needs were very different from those of the scientists and researchers who had previously been the network's primary users. Working in parallel with BBN Systems & Technologies, which ran the first Network Information Center (NIC) for the NSFNET, the Merit Information Services staff created an array of materials designed to help new users of the NSFNET from all areas of academia, who were not as experienced with networking. Merit was also instrumental in teaching the majority of today's Internet "trainers," who provide training materials and resources to help people learn how to use the Internet.
The Merit/NSFNET Networking Seminars, "Making Your Internet Connection Count," brought together Internet experts, network developers and administrators, information support specialists, librarians, and teachers who needed to know the latest about networking for their institutions. Beginning in 1989, seminars were presented on-site in conjunction with NSFNET regional networks in Hilton Head, Denver, Ann Arbor, Washington, D.C./College Park, Las Vegas, San Francisco, and Orlando.
Merit Information Services staff worked successfully with many institutions to design network training on a variety of topics, from library resources to discipline-specific databases. Merit's Advanced Topic workshops educated technical personnel from the NSFNET backbone site locations about advancements in Internet engineering, network management, and routing.
Merit published the Link Letter (ftp://nic.merit.edu/nsfnet/linkletter/) newsletter to keep the community informed about changes to the backbone and the broader Internet infrastructure. A presentation called "Cruise of the Internet," written by Merit's Laura Kelleher, introduced new and experienced Internet users worldwide to information resources, and introduced users to tools such as telnet, FTP, e-mail, Gopher, and WAIS. A popular, computer-based Cruise was developed by Steve Burdick, Kelleher, and Mark Davis-Craig for Macintosh and Windows users.
Kathleen McClatchey and other Merit staff distributed over 1,000 copies of a video produced in conjunction with EDUCOM's Networking and Telecommunications Task Force (NTTF), "Networking to Share Ideas," containing vignettes of NSFNET users using the communications potential of the network in such diverse areas as health care, environmental research, and K-12 education. In addition, the partners produced many technical working papers and articles about networking and the research and education community in magazines and other print media.
Merit, IBM, MCI, ANS, and the NSF also maintained a high profile for the project by attending and giving presentations about the NSFNET to national, regional, and local conferences and Internet-related organizations, including the Internet Engineering Task Force (IETF), EDUCOM, FARNET, the Coalition for Networked Information (CNI), and many others. Team members also published dozens of technical papers, including Internet standards documents called RFCs (Request For Comments), that helped push the leading edge of networking technology in the Internet community. Efforts to publicize the achievements of the NSFNET backbone service began to dovetail with the larger movement toward a National Research and Education Network, or NREN. Articles in as diverse publications as Science, the Wall Street Journal, and USA Today helped to raise awareness of the NSFNET beyond the research and education community. Ellen Hoffman remembers the difficulty in getting people "up to speed" with the concept of networking and the Internet:
"Now you just say Internet and the press gets all excited. It wasn't true back then; invariably if you wanted to talk about the Internet, since nobody knew what it was, you would have to go through a long process of explaining. You have to remember that back then you didn't have something like the Web that made it easy to show--we had e-mail, FTP, and telnet. Those in and of themselves are very interesting applications and they do some wonderful things functionally, but they are not very showy."
This marked the beginning of an educational process for journalists and their readers and listeners that would saturate the media with stories about the "information superhighway" by 1994.
More substantively, the partners had to come up with new ways of "making the NSFNET real to people." Focusing on how users incorporated networking technology into their daily work (and increasingly, other areas of their lives), on the innovative applications being developed, and how using the NSFNET was changing the ways people related to one another seemed the best strategy. "We were trying to show the incredible power in sharing information--the idea of the collaborative work enabled by and through the network," recalls Kathleen McClatchey. "There is so much we can do with the information that we have to make people's lives better."
Today, for the growing millions of people using the Internet and commercial online services, networking is increasingly as ubiquitous as the telephone. How that happened is a testament to the technology and, more importantly, to the people using it: technology is a social construct, and often the ways in which we think about it when it is created begin to change as a result of human innovation. "I remember when we were first installing computers on campus; people installed them for one reason and ended up using them for another reason, and often they didn't really know what that other reason was going to be at the time they purchased them," says Priscilla Huston. On the ARPANET, e-mail among researchers was more of an unexpected side benefit than the stated or primary purpose of creating a secure communications system. In a similar manner, what began as primarily a research network evolved to a broader backbone network service in support of the entire research and education community, and by the early 90s was evolving yet again in the context of public discussion about a National and eventually Global Information Infrastructure. The NSFNET was one of the main catalysts for these ideas. "We wouldn't be talking about the NII without the NSFNET," says Ellen Hoffman. That particular achievement is one of the NSFNET's proudest.
NSFNET team members agree on the primary factors that led to the success of this project: vision, flexible yet firm management, and commitment from individual staff at every level of the partnership. Over and over again, staff from Merit, IBM, MCI, ANS, and the NSF emphasized how important it was to have a clear objective guiding their way, and that the foundation for that goal--building a national backbone network--was a vision of what it could mean to people using it. Doug Van Houweling puts it this way:
"We all shared a sense of what this could be, and when the going got tough on getting the circuits to work or when people questioned our competence or motives, we fell back on our conviction 'This is Important, This is Significant.' A lot of people look for an opportunity to have a real impact on the world. When the chance arrives, it provides powerful motivation."
Clear objectives, however, didn't mean a rigid adherence to what had been done before, or to only one notion of how to do it. In a dynamic environment where, as Hans-Werner Braun noted, expectations and events would change on a day-to-day basis, flexibility was key. "Disguised limits, no preconceived limitations," is how Mathew Dovens puts it. "This project was set up to open new horizons, and I think throughout the seven years we never limited ourselves." Flexibility was balanced with firm guidance, and by all accounts Merit and the NSF provided just the right mix of incentives. Harvey Fraser notes particularly the NSF's contribution:
"Steve [Wolff] had very high goals, and did not interfere with implementation, but many times there were choices to be made which may or may not match the most recent proposal that's been funded. In the process of exploring those choices, it always came out that what he wanted was technological leadership and excellence of operation, and however that got manifested in a given situation, it didn't matter. If it meant that there was a need for the government to invest more money, buy more circuits, build something else, whatever it was, fine. He certainly provided a very steady hand on the tiller."
Other team members singled out Eric Aupperle as having managed the building of the NSFNET backbone well: "Eric was always a gentleman-- he was forceful in expressing his views, but was always ready to find a compromise."
Perhaps the most important contributor to the success of the NSFNET project, however, was the commitment among the partners--Merit, IBM, MCI, ANS, and the NSF--at every level, to make it come together.
All of the NSFNET team stressed the importance of "buy-in" to the project on the executive, managerial, and technical staff levels of the partnership. Part of this was the individual commitment to the overall "vision" of the NSFNET. But a willingness on the part of each person to commit to the job before them was necessary as well, even if it meant long hours, excruciatingly difficult technical problems, or criticism from the community.
Another aspect of this commitment showed in the partners' willingness to collaborate closely throughout the length of the project, through a series of technical and administrative meetings that took place on a weekly, monthly and quarterly basis. Finally, in the back of their minds, the NSFNET team members knew that if they didn't succeed, there was no other alternative for the community. According to the NSF and Merit, they recognized that as the NSFNET was at first the only backbone network, they couldn't let the community down and were obligated to put the very best network in place. Mark Knopper says:
"This was it: there was no other backbone. There was no competition. The user community was there, demand for the bandwidth brought in the traffic, the regionals aggregated it, we had to carry it. That forced us to look ahead, time and again, because no one else was in a position to do it."
Immense personal satisfaction came from working as a team where commitment, both to the project and to a larger vision, flexibility, and skilled management were part of the environment around them. "To make it happen and really make a difference, and to drive the whole thing forward," says Hans-Werner, was the most rewarding. Mark Knopper speaks for them all: "I was honored by working with those people."
What lessons can be learned from this successful partnership between the research and education community, industry, and government? Could the nation do it again? "I think it was unique, but it doesn't have to be," says Elise Gerich. Mark Knopper emphasizes the importance of NSF funding as well as industry cost-sharing; according to him and others, "leveraging was a really, really big win" for the project. Steve Wolff thinks that similar initiatives would benefit from "finding out when the interest of the academic community and industry are similar, and forming a partnership with government participating at the margin. It's a great model," he says.
Also important was the way in which the NSF and Merit could bring different players together: "A central element of the success was the extremely productive interaction between industry, academia, and government," says Paul Bosco. Walter Wiebe notes that "being part of the team that pulled it together will be a source of satisfaction to me all my life." According to Priscilla Huston, "the NSFNET project has involved a lot of different parties, not just the NSF, but international and industry players. Merit's support of that was an important part of the award."
Last but not least, the way in which the funding was applied, the overall construction of the NSFNET program--of which the backbone service was the key piece--was done in a way that would reap the maximum benefits from infrastructure building. "You really have to be building at both levels at the same time," explains Ellen Hoffman.
"In creating a ubiquitous network, you have to do both. You want it to be widespread--first of all the NSF built a backbone with lots of bandwidth, a playground for people to work with. At the same time they were doing other things to start making it widely available, like the Connections program; almost every college in this country started off with an NSF grant for its first connection. That is the other part of the ubiquity. And they didn't want to do only one or only the other, because they recognized that if they hadn't built the backbone we wouldn't have resources like the Web."
All of these factors combined, say the partners, helped to make the NSFNET a success, to build the very best national backbone network of its kind at the time. Says Al Weis: "What the NSF accomplished was outstanding. They took a little bit of seed money and created the right dynamics which, in essence, sparked the Internet, the global Internet."
By the summer of 1991, as the NSFNET partners continued to work on the T3 upgrade, the NSF knew that the five-year cooperative agreement with Merit would be ending in 1992. Around this time, according to George Strawn, NCRI began thinking and planning in earnest about what to do next. It seemed clear that in the context of the debate about commercialization, continuing on the same course of support for the NSFNET was not an option; nor was the NSF particularly interested in doing so. Steve Wolff saw the actions of the NSF with the NSFNET as similar to the NSF's support of campus computing centers in the early sixties and seventies, which was phased out after computing products and services were readily available from the marketplace. The NSF knew in both cases when to initiate high-technology initiatives, and when to get out. Wolff explains the NSF's philosophy in this way:
"I think it's a strength. You're subject to criticism, because people do get hurt in the process; there's no question about that. On the other hand, NSF recognizes limitations and only has so much money. If you don't stop doing old things, then you can't start any new things. And when something gets to the point that it becomes a commodity product, there is no reason for NSF to be supporting it."
Internet service providers were springing up all over the country, from local dial-up providers to larger companies providing T1 and eventually T3 service, and there were now a number of vendors offering TCP/IP networking products and services. During 1992, the National Science Board authorized an extension of Merit's cooperative agreement for eighteen months beyond the October 1992 expiration date in order for NCRI to develop the solicitation. Recognizing that the best way to solve some of the thorny issues precipitated by the pending termination of the cooperative agreement--and the commercialization question--was to involve the community in its plans more directly, in the early summer of 1992 the NSF distributed a draft of its solicitation for a new architecture to the public. The design for the new architecture was based on concepts developed in a paper titled "NSF Implementation Plan for Interim NREN" by Bob Aiken of the NSF; Hans-Werner Braun, then at the San Diego Supercomputer Center; and Peter Ford, on assignment to the NSF from Los Alamos National Lab.(5)
The draft solicitation was written by George Strawn, then NSFNET Program Director under Wolff at NCRI. "Steve said, 'This is really complicated, and it will be really important for us to get it right. I think we need to put out a public draft,'" remembers Strawn. NSF received over 240 pages of comments on the draft; Peter Ford and NSF's Don Mitchell helped Strawn incorporate the comments into the final plans. Finally, in the spring of 1993 the NSF released Solicitation 93-52, Network Access Point Manager, Routing Arbiter, Regional Network Providers, and Very High Speed Backbone Network Services Provider for NSFNET and the NREN Program.
The solicitation contained four parts: a very high-speed backbone, or vBNS; Network Access Points; a Routing Arbiter; and Regional Network Provider awards. The vBNS, linking five NSF supercomputer centers and operating at minimum speeds of OC-3 (155 Mbps), would offer a high- speed, high-bandwidth virtual infrastructure for networking research, including advanced applications. Network Access Points, or NAPs, would act as interconnection points for commercial Internet service providers. The Routing Arbiter would manage the ever-growing routing tables and databases for the providers connecting at the NAPs. Finally, regional and midlevel networks would continue to receive NSF funding, phased out over a four-year period, to support their connections to the Internet. They would use the money to pay commercial Internet service providers, who were then required to connect to NAPs. In this way, the NSF hoped, connectivity for the research and education community to the rest of the Internet would be maintained.
During 1993, as the responses to the solicitation came in to the NSF, the partners continued to maintain the NSFNET backbone's stable networking environment. Various hardware and software improvements to the backbone were added, and a major router upgrade doubled the speed of packets switched on the network. Merit also introduced its new Policy Routing Database, or PRDB, an expansion and improvement of routing procedures used on the NSFNET backbone service since its inception. At the same time, Merit prepared and eventually won its proposal for a Routing Arbiter award. All of the winners of awards for the new architecture were announced throughout 1994. Merit and the Information Sciences Institute (ISI) at the University of Southern California together form the Routing Arbiter team.
The vBNS award was given to MCI; Network Access Point Manager awards were given to Sprint, for a New York NAP; MFS Datanet for a Washington D.C. NAP; and Bellcore, which manages two NAPs: the Chicago NAP in conjunction with Ameritech, and the California NAP in conjunction with Pacific Bell. Awards to the regional and midlevel networks for interregional connectivity were given to seventeen regional and midlevel networks.
In May of 1994, Merit's cooperative agreement was extended one last time through April of 1995. From 1994 through the first half of 1995, Merit and NCRI worked closely with the regional and midlevel networks to ensure a smooth transition, helping them to disconnect from the NSFNET backbone service: "Merit's award was winding down, but it was their responsibility to get people moved over and they did a really good job pulling people together," says Huston. The PRDB was also succeeded by the Routing Arbiter Database, a global registry of routing and networking information that forms an important part of the new Internet Routing Registry.
MCI began to operate the vBNS as well as provide networking services to many of the NSFNET's regional and midlevel networks through internetMCI. IBM formed Advantis, the US-based part of the IBM Global Network, with many of the personnel from the NSFNET project. The NSFNET partners also continued to provide T3 service to the NSF- sponsored supercomputer centers until the vBNS came online. On April 30, 1995, the NSFNET backbone service was decommissioned, and an era ended--in triumph.
The NSFNET backbone service will be remembered not only for achieving the goal of providing networking connectivity to the research and education community, but also for realizing some of the larger hopes of the partners: that networking would become a ubiquitous part of everyday life for more and more people, and that it would create new markets for products and services in networking and communications, thus promoting technology transfer.
Merit, IBM, MCI, ANS, and the National Science Foundation knew that their immediate task was to improve on the previous NSFNET backbone, to push the technology past what had ever been attempted before, and to implement it on a national scale. In doing so, they hoped to place a powerful tool in the hands of the research and education community, and spur innovation in the development and use of communications infrastructure and applications. The partners were successful on both counts: the NSFNET backbone service connected most of the higher research and education community in the US to a robust and reliable high-speed networking communications infrastructure. It also played a singular role in creating an Internet "industry," a dynamic new marketplace for communications technologies. As the NSFNET spurred the growth of the Internet, moreover, use of networking technology expanded outward from scientists and educators in academia and industry to encompass individuals in all kinds of enterprises and organizations, both private and public, all in a few short years.
The NSFNET program will continue to evolve, to reflect the changing needs of the research and education community as well as the revolution in communications infrastructure it had such a large hand in creating. Today's new architecture does not represent a "new" NSFNET. Rather, the NSF sought awardees to provide the main structural elements of the next generation of high-speed networking in the US as well as internationally. To accomplish these goals, the NSFNET program is moving in two directions: to provide support for the research and education community by ensuring the availability of services, resources and tools necessary to keep the larger Internet interconnected, and to continue to push the envelope of networking technology with projects such as the vBNS--all in the context of the burgeoning market for commodity providers of Internet services and applications.
The success of the NSFNET project demonstrates the potential of partnerships between academia, business, and government in high- technology initiatives. The keys to its success seem to lie in the combination of vision, technical achievement, commitment, and luck that marked the project throughout its seven and one-half years of operation.
The partners began with a vision of what the NSFNET backbone service could mean to the research and education community, enabling them to work collaboratively, no matter where they were physically located, and share the results of their intellectual productivity in seconds. This vision was supported by commitment to achieving the project goals at all levels of the organizations involved: by all of the programming and engineering, information and user services, managerial and executive staff at Merit, IBM, MCI, ANS, and the National Science Foundation. An important part of that vision was the philosophy of building infrastructure at two levels simultaneously: constructing the NSFNET backbone service at the top and providing for regional network connectivity while empowering individuals at the base by funding campus connections to the NSFNET. Another important part was to expand the original notions of research networking to include education, following the example set by modern research universities that attempt to combine teaching expertise with research activity.
The partners reached the technology frontiers of networking and communications and crossed them--from building a nationwide T1 backbone service to reengineering, expanding and upgrading a nationwide T3 backbone service that at its peak transferred over 86 billion packets per month at 45 Mbps speeds--the equivalent of the contents of the Library of Congress every two weeks. At the same time, the NSFNET partners also had to respond to daunting challenges, both technical and otherwise, engendered by building networking infrastructure on a national scale: as the NSFNET (and the Internet) exploded in popularity, a complete upgrade to T3 speeds was required to support increased use, which in turn provoked questions of making the technology self-supporting by commercializing the technology. The transition from the NSFNET backbone service to a new architecture incorporating commercial Internet service providers resulted from the partners' creation of ANS and involvement in community discussion of issues of "commercialization and privatization" of the Internet. All of these elements accounted for the project's success, together with a bit of good fortune: "I don't think there's any substitute for luck," Doug Van Houweling says. We can learn much from the example set by the NSFNET.
Since the earliest days of the telegraph and the telephone, history tells us that the arrival of each new communications medium has been accompanied by grandiose claims of its potential benefits to society. In order to take advantage of the exciting opportunities afforded by today's technology, it is imperative that policy makers examine the development of the NSFNET and the Internet. We are still far away from a truly open, interoperable, and ubiquitous global information infrastructure accessible to all, "from everyone in every place to everyone in every other place, a system as universal and as extensive as the highway system of the country which extends from every man's door to every other man's door," in the words of Theodore Vail, president of AT&T in 1907. However, the Internet has brought us a giant step closer to realizing the promise of high-speed networking, one of the most revolutionary communications technologies ever created. As part of this phenomenon, the NSFNET backbone service provided a model for future partnerships as well as a legacy of technology for the world.
(1) Dennis Jennings, Laurence Landweber, Ira Fuchs, and David Farber, "Computer Networking for Scientists," Science 23 (February 1986).
(2) EDUCOM, Networking and Telecommunications Task Force, "A National Higher Education Network: Issues and Opportunities." May 1987.
(3) Office of the Inspector General, "Review of NSFNET." 1993: National Science Foundation 93-01, p. 38.
(4) Office of the Inspector General, "Review of NSFNET." 1993: National Science Foundation 93-01, p.68.
(5) B. Aiken (NSF), H.-W. Braun (SDSC), and P. Ford (LANL); ed. K. Claffy (SDSC), "NSF Implementation Plan for Interim NREN" (May 1992), Journal of High Speed Networking, Vol. 2, No. 1, 1993.
This report was researched and written by Karen D. Frazer.
Merit Network, Inc. would like to thank all those who contributed to the report: Neal Lane, Ellen Hoffman, Mike Roberts, Jane Caviness, Priscilla Huston, Doug Van Houweling, Dale Johnson, Harvey Fraser, Mark Knopper, Elise Gerich, Al Weis, Eric Aupperle, Don Mitchell, Steve Wolff, Walter Wiebe, Mathew Dovens, Jack Drescher, Paul Bosco, Larry Bouman, Hans-Werner Braun, Dan van Bellegham, Bob Mazza, Rick Boivie, Alan Baratz, Kathleen McClatchey, Jessica Yu, and George Strawn.
Many dedicated people helped plan, build, operate, and promote the NSFNET backbone service. It is hoped that this report will serve to recognize the many team members, named and unnamed, who helped to build the NSFNET.
This material is based upon work supported by the National Science Foundation under Grant No. NCR 8720904. The Government has certain rights to this material. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.