The Launch of NSFNET
While CSNET was growing in the early 1980s, NSF began funding improvements in the academic computing infrastructure. Providing access to computers with increasing speed became essential for certain kinds of research. NSF's supercomputing program, launched in 1984, was designed to make high performance computers accessible to researchers around the country.
The first stage was to fund the purchase of supercomputer access at Purdue University, the University of Minnesota, Boeing Computer Services, AT&T Bell Laboratories, Colorado State University, and Digital Productions. In 1985, four new supercomputer centers were established with NSF supportthe John von Neumann Center at Princeton University, the San Diego Supercomputer Center on the campus of the University of California at San Diego, the National Center for Supercomputing Applications at the University of Illinois, and the Cornell Theory Center, a production and experimental supercomputer center. NSF later established the Pittsburgh Supercomputing Center, which was run jointly by Westinghouse, Carnegie-Mellon University, and the University of Pittsburgh.
In 1989, funding for four of the centers, San Diego, Urbana-Champaign, Cornell, and Pittsburgh, was renewed. In 1997, NSF restructured the supercomputer centers program and funded the supercomputer site partnerships based in San Diego and Urbana-Champaign.
A fundamental part of the supercomputing initiative was the creation of NSFNET. NSF envisioned a general high-speed network, moving data more than twenty-five times the speed of CSNET, and connecting existing regional networks, which NSF had created, and local academic networks. NSF wanted to create an "inter-net," a "network of networks," connected to DARPA's own internet, which included the ARPANET. It would offer users the ability to access remote computing resources from within their own local computing environment.
NSFNET got off to a relatively modest start in 1986 with connections among the five NSF university-based supercomputer centers. Yet its connection with ARPANET immediately put NSFNET into the major leagues as far as networking was concerned. As with CSNET, NSF decided not to restrict NSFNET to supercomputer researchers but to open it to all academic users. The other wide-area networks (all government-owned) supported mere handfuls of specialized contractors and researchers.
The flow of traffic on NSFNET was so great in the first year that an upgrade was required. NSF issued a solicitation calling for an upgrade and, equally important, the participation of the private sector. Steve Wolff, then program director for NSFNET, explained why commercial interests eventually had to become a part of the network, and why NSF supported it.
"It had to come," says Wolff, "because it was obvious that if it didn't come in a coordinated way, it would come in a haphazard way, and the academic community would remain aloof, on the margin. That's the wrong modelmultiple networks again, rather than a single Internet. There had to be commercial activity to help support networking, to help build volume on the network. That would get the cost down for everybody, including the academic community, which is what NSF was supposed to be doing."
To achieve this goal, Wolff and others framed the 1987 upgrade solicitation in a way that would enable bidding companies to gain technical experience for the future. The winning proposal came from a team including Merit Network, Inc., a consortium of Michigan universities, and the state of Michigan, as well as two commercial companies, IBM and MCI. In addition to overall engineering, management, and operation of the project, the Merit team was responsible for developing user support and information services. IBM provided the hardware and software for the packet-switching network and network management, while MCI provided the transmission circuits for the NSFNET backbone, including reduced tariffs for that service.
Merit Network worked quickly. In July 1988, eight months after the award, the new backbone was operational. It connected thirteen regional networks and supercomputer centers, representing a total of over 170 constituent campus networks and transmitting 152 million packets of information per month. Just as quickly, the supply offered by the upgraded NSFNET caused a surge in demand. Usage increased on the order of 10 percent per month, a growth rate that has continued to this day in spite of repeated expansions in data communications capacity. In 1989, Merit Network was already planning for the upgrade of the NSFNET backbone service from T1 (1.5 megabits per second or Mbps) to T3 (45 Mbps).
"When we first started producing those traffic charts, they all showed the same thingup and up and up! You probably could see a hundred of these, and the chart was always the same," says Ellen Hoffman, a member of the Merit team. "Whether it is growth on the Web or growth of traffic on the Internet, you didn't think it would keep doing that forever, and it did. It just never stopped."
The T3 upgrade, like the original network implementation, deployed new technology under rigorous operating conditions. It also required a heavier responsibility than NSF was prepared to assume. The upgrade, therefore, represented an organizational as well as a technical milestonethe beginning of the Internet industry.
In 1990 and 1991, the NSFNET team was restructured. A not-for-profit entity called Advanced Networks and Services continued to provide backbone service as a subcontractor to Merit Network, while a for-profit subsidiary was spun off to enable commercial development of the network.
The new T3 service was fully inaugurated in 1991, representing a thirtyfold increase in the bandwidth on the backbone. The network linked sixteen sites and over 3,500 networks. By 1992, over 6,000 networks were connected, one-third of them outside the United States. The numbers continued to climb. In March 1991, the Internet was transferring 1.3 trillion bytes of information per month. By the end of 1994, it was transmitting 17.8 trillion bytes per month, the equivalent of electronically moving the entire contents of the Library of Congress every four months.