Subject: N-1-4-010.10 Changing Eras: Evolution of the NSFNET Eric M. Aupperle On Wednesday, 2 December 1992 the T1 NSFNET service ended. All NSFNET traffic now traverses ANS's T3 backbone infrastructure. The transition of traffic transport from the T1 to the T3 backbone service occurred gradually, most notably since March of 1992. As this T1 era ends and a new one advances, it seems appropriate to reflect on NSFNET's evolution. The genesis of NSFNET dates back to 1985 when the National Science Foundation committed support for helping establish five new supercomputer facilities. With their objective of sharing these supercomputer resources among the nation's research universities, NSF required each supercomputer site to provide network access to a set of participating institutional partners. An example of this was the San Diego Supercomputer Center's 1986 implementation of a 56 Kbps American Satellite system linking distant organizations to their site. NSF also took other steps during this period to broaden access to these new advanced computing facilities. They arranged for shared use of existing national community data networks, i.e., ARPAnet and BITNET. They funded the National Center for Atmospheric Research's deployment of the University Satellite Network Project to interconnect several universities across the United States. They encouraged the creation of new regional network organizations to provide yet greater connectivity to the educational and research community, and partially funded them. Another key NSF network action occurred in 1986 when they contracted for the first NSFNET backbone. With the communications access facilities the supercomputer centers had in place, this backbone's mission focused on providing inter supercomputer site connectivity to foster more national resources sharing opportunities. It's role soon emerged as a network of networks, at the top of a three-level hierarchy. This initial backbone linked six sites with 56 Kbps data circuits. It quickly became saturated with traffic, experienced routing difficulties, and lacked the funding and organizational resources to address these issues. At it's peak this backbone transported about 115 million packets per month during the first half of 1988. A 1987 NSF solicitation process led to an award to Merit Network, in partnership with IBM and MCI, for managing, operating and continuing the development of the NSFNET backbone. This solicitation required providing connectivity for thirteen sites, the six supercomputer centers and seven regional networks. In July of 1988 the Merit partnership replaced the initial 6 node 56 Kbps network with a with 13 node network based on T1 data trunks. These 1.5 Mbps T1 data circuits were multiplexed to provide multiple 500 Kbps links among the 13 nodes. With the new backbone and rapidly expanding regional network infrastructure, backbone traffic spurted, averaging 20% per month of growth between July 1988 and July 1989. By mid 1989 the Merit partnership added new T1 circuits and began operating all of them as full 1.5 Mbps links in response to this growth. T1 connections also were installed for two interagency network connection sites named FIX East and West, and later a fourteenth node was added at NSF's request. Backbone traffic continued growing rapidly as did the number of announced networks supported on the backbone. These factors drove the implementation of the T3 network service which now fully replaces the T1 network. The T1 traffic peaked in February 1992, with a load of about 11.3 billion packets for that month. This represented a two order of magnitude increase of packet traffic over the network it replaced. It's time had come and gone in less than four years. Meanwhile the relentless increase in backbone traffic continues. This past November saw the first occurrence of more than a billion packets shipped during a single day. The total November monthly traffic was nearly 24 billion packets. Any bets on the lifetime of the new era?