packet-switched network
packet-switched network, digital network that transfers data files by breaking them up into smaller units (“packets”) and transmitting them through a series of nodes (“switches”). In a process known as store-and-forward, the switches route the packets to their destination, where the data files are then reassembled. The Internet and most local area networks (LANs) are packet-switched networks.
Definition
Traditional networks, known as circuit-switched networks, transmit data through a single physical path of fixed bandwidth, holding the path open for as long as necessary. All information is sent down the path in sequential order. Packet-switched networks, on the other hand, can direct packets of information down multiple paths; each “switch” is made based on improving efficiency. Additionally, if outages were to occur on a certain path, the packets would be rerouted down other paths. Consequently, packets might arrive out of order and must be reassembled at their destination.
Packet-switched networks have certain benefits over circuit-switched networks, such as optimized channel capacity and improved fault tolerance. However, packet switching is relatively complicated and thus requires significant processing power and large amounts of random-access memory (RAM). Packet switching can also involve rerouting delays and packet loss. Packet switching is therefore preferred for transmitting relatively small files, while circuit switching is still used for larger transfers (e.g., real-time applications).
Packet-switched networks have two main components: cores, which consist of routers and control systems connected by high-bandwidth communication channels, and edges, which are where end-user systems are located. Hosts are end-user systems (e.g., personal computers) capable of sending and receiving data packets. Communication across network cores occurs via protocols, which are procedures employed by senders and receivers to communicate effectively. The protocols used to transfer end-users’ data are collectively referred to as a protocol stack.
Each transmission packet of data (“datagram”) that passes through a packet-switched network’s core is made of a header and payload. The header contains control information, such as the addresses of the packet’s sender and receiver. The payload consists of the information to be delivered. Packets are sometimes split into even smaller units in a process known as packet fragmentation.
Packet-switched networks take the form of connectionless networks (also known as datagram networks) and connection-oriented networks (also known as virtual-circuit networks). Connectionless networks work as described above: data is partitioned into smaller pieces, which are then attached to formatted headers, creating datagrams. The datagrams take the best possible route from the source to the destination. Connection-oriented networks, on the other hand, mimic circuit-switched networks by setting up a dedicated route between the sending host and the receiving host prior to packet transfer, thus gaining the benefits of circuit-switching while on digital networks.
History
The idea to break up large units of data into smaller packets was first imagined by Paul Baran, an engineer at the RAND Corporation. Baran came up with what he called “hot-potato routing” to answer the U.S. Air Force’s question of how a computer communications network could continue to function through a nuclear attack. He published his theory first in a series of RAND studies between 1960 and 1962 and then in an 11-volume analysis titled On Distributed Communications in August 1964. At the time the work garnered little interest from the government or private corporations.
Meanwhile Donald Davies, a computer scientist at the United Kingdom’s National Physical Laboratory (NPL), independently came up with the same concept and began working on a network on which to test the idea. While Baran called his units of data “message blocks,” Davies referred to his units as “packets.” Lawrence (Larry) Roberts, a manager at the U.S. government’s Advanced Research Projects Agency (ARPA; now Defense Advanced Research Projects Agency [DARPA]), learned of Davies’s work in October 1967 at a symposium in Gatlinburg, Tennessee. Roberts adopted Davies’s term packet switching for ARPANET, his forthcoming project.
ARPANET was the first public packet-switched network. The first successful test occurred in October 1969, and the network was designed in the span of a year by the technology consultancy company Bolt Beranek and Newman (BBN; now Raytheon BBN), which referred to Baran’s and Davies’s concepts and designs. Initially, the network connected nodes in only four locations: the University of California at Los Angeles (UCLA), Stanford Research Institute, University of California at Santa Barbara, and the University of Utah. By 1975 ARPANET had 57 nodes.
ARPANET was showcased in October 1972 at the first International Conference on Computer Communications (ICCC). The presentation showed that it was possible to create a packet-switched network, but much of the communications industry in the United States remained uninterested, if not outright hostile. Seeing an opportunity, BBN and Roberts founded the commercial network company Telenet that year.
Some telecommunications agencies in other countries were also bullish about the new technology. In November 1973 the French postal and telecommunications division announced a domestic packet network named TRANSPAC. The Trans-Canada Telephone System announced its own DATAPAC in October 1974. In Japan the Nippon Telegraph and Telephone Corporation planned to create its own public packet-switched data network as well. Most providers remained skeptical, preferring to wait and see how the early public networks performed.
While those public networks were under construction, researchers developed new packet-switched networks that experimented with and improved on the packet-switching concept. After finishing his Mark I network at the NPL in 1970, Davies completed the Mark II network, a project that would gradually influence the United Kingdom and much of Europe, in 1973. That same year French computer scientist Louis Pouzin finished CYCLADES, a datagram-based network that greatly influenced computer communications by making hosts, rather than network cores, responsible for error correction. The internationally funded European Informatics Network (EIN) came online in 1976.
There was strong motivation for the five nations building public packet networks—Canada, France, Japan, the U.K., and the U.S.—to agree on a standard host-network interface, so talks on the subject began in 1975. The result was the protocol CCITT Recommendation X.25, which was adopted by all of the nations involved in March 1976. X.25 ushered in the next phase of packet switching: the rise of interconnected public service networks. Additional agreements soon followed, among them X.75, a standard protocol for connecting international networks.
In 1979 Robert (Bob) Kahn became director of the Information Processing Techniques Office (IPTO) of DARPA. By this time the U.S. Department of Defense had multiple packet-switched networks, but none of them were compatible with each other. Kahn united the systems by having the Defense Department adopt TCP/IP, a protocol standard that Kahn had first imagined in a 1974 paper he had written with the prominent software developer Vincent Cerf. Use of TCP/IP spread to other research laboratories and to the public at large, eventually becoming the basis for the ultimate packet-switched network: the Internet.