Internet Guide Logo

NSFNET

bullet Introduction

NSFNET was a computer network funded by the National Science Foundation (NSF), which is an agency of the government of the United States of America. NSFNET stands for the 'National Science Foundation Network', and from 1987-1992, NSFNET was viewed as the backbone network/infrastructure of the Internet. The original NSFNET backbone consisted of six supercomputer locations, which were located at U.S. universities and research institutions. NSFNET used the TCP/IP protocol suite to connect and transport data across its network; the Internet is a term coined by the computer scientists who designed TCP/IP and described the ability of TCP/IP to interconnect computer network into one large super-network. In 1987, Ed Krol wrote the Hitchhiker's Guide, which was the first popular guide for the Internet, and gave a detailed description of NSFNET; this book is now republished as RFC 1118.

bullet NSFNET evolves from ARPANET and CSNET

ARPANET was the first computer network in the United States to connect research and education institutions. ARPANET pioneered packet switching, and the TCP/IP protocol suite was invented by Vint Cerf and Bob Kahn, in the 1970s, as an improvement to ARPANET host-to-host protocol program. Due to the Mansfield Amendment of 1973, the Department of Defense (ARPA was a US defense agency) was directed to no longer fund the development of science projects. ARPANET was split into a military and a research network, and the research network of ARPANET was planned to be slowly phased out from 1975 onwards.

Due to a lack of funding to expand and develop ARPANET, the National Science Foundation (NSF) funded the creation of CSNET (Computer Science Network). In 1979, Lawrence Landweber, Professor Emeritus of Computer Science at the University of Wisconsin, had proposed the creation of CSNET, with support from Dave Crocker, Vint Cerf and DARPA, to connect U.S. universities, industries and government computer research groups. His proposal received $5 million in funding in January 1980, and the team who built CSNET included: L. Landweber, B. Kern, P. Denning, and T. Hearn. One of the reasons CSNET was created was to connect U.S. research and education institutions who could not connect to ARPANET.

CSNET initially connected the following U.S. institutions (amongst others): University of Minnesota, University of Wisconsin, University of Oklahoma, University of California-Berkeley, and Yale University. From 1981-1984, CSNET was funded by the National Science Foundation, and by 1984 the network had been connected to over 150 universities. Due to the success of CSNET, the National Science Foundation planned the creation of a new expanded network in 1984: NSFNET. When NSFNET was launched 1986, it soon eclipsed ARPANET and CSNET, which were both slowly decommissioned from 1985-1991.

bullet NSFNET is planned

The National Science Foundation (NSF) funded a range of computer science projects: chief amongst them was the creation of a supercomputing program in the early 1980s. Science projects had increasingly become reliant on computers, and relied upon powerful computing power. Part of the NSF's supercomputing program was to connect the supercomputers together to form a computer network. This network would become NSFNET, and it was envisaged that data could travel up to twenty-five times faster on NSFNET than it could on CSNET: due to the power of the supercomputers. By 1984, the NSF had funded the construction of supercomputers at the following locations:

  1. Princeton University (John von Neumann Center)
  2. University of California, San Diego (Supercomputer Center)
  3. University of Illinois (National Center for Supercomputing Applications)
  4. Cornell University (Cornell Theory Center)
  5. University of Pittsburgh (Pittsburgh Supercomputing Center (PSC))
  6. National Center for Atmospheric Research (NCAR)

In 1985, the NSF hired Dennis Jennings to manage the creation of NSFNET, Jennings was an Irish physicist and computer networking pioneer who was an active participant in the creation of the TERENA European research network. Jennings decided that NSFNET would: function as an all-purpose research network; it would act as a backbone network that would facilitate the connection of regional networks to it; and it would use the TCP/IP protocol suite that had been implemented on CSNET, MILNET and ARPANET. The infrastructure plan, outlined by Jennings, for NSFNET was as follows: -Backbone (supercomputer centers); -Tier Two: Regional Networks; -Tier Three: Campus Networks.

The development of TCP/IP had been funded by DARPA, and it became a Department of Defense standard in the early 1980s. When NSF picked TCP/IP for its network, it had elected to support the DARPA Internet organisational infrastructure, such as the IAB, that managed the development of TCP/IP and its namespace resources. During the 1980s, there was a competition between the IP and OSI networking models, referred to as the protocol war, and it was feasible that Jennings could have used OSI instead of IP for NSFNET. Jennings decision was crucial for the development of IP networks, which would eventually interconnect to become the modern Internet in the early 1990's.

bullet NSFNET is operational

By the end of 1986, NSFNET was operational, and the team that built NSFNET included: members of NCSA (included Ed Krol and C. Kline); members of CUTC (included A. Brown and S. Brim); members of the University of Delaware (included D. Farber and D. Mills); and H.W. Braun of Merit. NSFNET was built upon a backbone of six supercomputer locations, operating at a data speed of 56 K-bit/sec, with each supercomputer connected to a newly designed 'fuzzball' router - the fuzzball tested new routing protocols like NTP.

The first problem that arose with the NSFNET backbone was its data speed: as more regional networks were connected, traffic congestion became an issue, and it became apparent that the NSFNET backbone infrastructure needed to be upgraded to support a faster data speed. In 1987, program director Steve Wolff solicited the private sector to upgrade the NSFNET infrastructure. In 1987, NSFNET signed a $39 million five-year cooperative agreement with Merit, IBM, and MCI to upgrade the NSFNET backbone. In 1988, the NSFNET backbone was upgraded to a 1.5 Mbit/s T1 network that featured thirteen nodes, and by 1991, it had been upgraded to a 45 Mbit/s T3 network that featured sixteen nodes. IBM focused on upgrading the packet switching hardware and software of NSFNET, and MCI upgraded the transmission circuits.

From 1987-1991, the NSFNET backbone was connected to a variety of regional and federal computer networks. To name but a few:

  1. BARRNet (Bay Area Regional Research Network)
  2. ESnet (Energy Sciences Network)
  3. MichNet (Michigan Network)
  4. MIDnet (Midwest Network)
  5. MILNET (Military Network)
  6. NSN (NASA Science Network)
  7. NorthWestNet (North West Network)
  8. NYSERNet (New York State Education and Research Network)
  9. SESQUINET (Sesquicentennial Network)
  10. SURAnet (South Eastern Universities Research Association Network)
  11. Westnet (West State Network)

Most of the regional networks listed above were connected to smaller (campus) networks, which numbered in the thousands. NSFNET was interconnected to other U.S. federal networks when the Federal Internet Exchange (FIX) was established in 1989. Therefore, for the first time, a 'network of networks' was formed, that would closely resemble the modern Internet. The NSFNET backbone was at the 'heart' of this configuration, and became known as the Internet's backbone. By 1992, over 4,000 networks in the United States, and over 2000 international networks, were connected to the NSFNET backbone.

bullet NSFNET and commercialisation

NSFNET was designed to foster communication between educational and research institutions in the United States of America. Therefore, the NSF, generally speaking, did not allow commercial use of its backbone network. By 1991, commercial networks (Internet Service Providers) were being founded to provide public/commercial access to IP networks. NSFNET created an 'Acceptable Use Policy' to allow these networks to connect to the NSFNET backbone. The problem was that the 'Acceptable Use Policy' of NSFNET did not allow unrestricted commercial use. In 1991, three commercial computer networks decided to create the Commercial Internet Exchange (CIX) : CIX would allow commercial traffic to be exchanged between commercial networks, which included:

  1. CERFnet (California Education and Research Federation Network)
  2. PSInet (Northern Virginia Network)
  3. UUnet (UUNET Communications Services)

At the same time, Merit, IBM, and MCI created a new commercial Internet Service Provider, named ANS CO+RE, that used the network infrastructure designed for NSFNET. Therefore, a situation arose where the network hardware of NSFNET was being used for two purposes: the NSF backbone service and the commercial traffic of ANS CO+RE. Another problem arose when ANS CO+RE refused to interconnect their network with the Commercial Internet Exchange (CIX): which created a perceived monopoly of the network hardware of the NSFNET backbone. While a compromise was eventually reached between ANS CO+RE and CIX, the controversy created by the situation would lead to a reorganisation of the Internet's backbone.

Instead of a centralised Internet backbone, that was federally funded by the U.S. government, it was decided that the backbone of the Internet would consist of commercial networks - typically large telecom companies - who would charge people to access their network. The transition from the NSFNET backbone to a privatised commercial backbone was a drawn out process, involving Network Access Points (NAP), that exchanged data between these commercial backbone networks. NAPs were the forerunner of Internet Exchange Points; such as The London Internet Exchange (LINX). The NSFNET backbone was officially decommissioned on the 30th of April 1995. NSFNET was instrumental in the development of the modern Internet.