|
The ARPANET Before the Internet, there was the ARPANET (1969-1990). The Advanced Research Projects Agency (ARPA) of the U.S. Department of Defense (DoD) created the ARPANET as an experiment in packet-switched computer networking. Although the ARPANET was the precursor to the Internet, the two are not the same. The first protocols used on the ARPANET were not TCP/IP; rather the TCP/IP protocols were invented using the ARPANET. The ARPANET, however, was the first transcontinental backbone network in the Internet. Decommissioned in 1990, it was made obsolete by the newer network technologies it in part helped create. |
|
ARPANET Services ARPA was funding supercomputers at a few research institutions and had many requests for more supercomputers at other locations. Rather than buying more supercomputers, the government considered a network that could permit researchers to access supercomputers remotely might be more cost-effective. The first four nodes of the ARPANET were communicating with each other by 1969. Resource sharing, in the form of remote login and file transfer, was thus one of the earliest goals of the ARPANET. |
Surviving a Nuclear War
The ARPANET was a direct product of the Cold War. ARPA itself was formed in response to the launching of the Sputnik satellite by the Soviet Union in 1957. It was intended to ensure the U.S. got ahead in military research and stayed there. As DoD urgently wanted military command and control networks that could survive a nuclear war, ARPA was charged with inventing a technology that could get data to its destination reliably even if arbitrary parts of the network disappeared without warning as a result of a nuclear attack. Traditional telephone technology, called circuit switching, was considered too fragile for the purpose, because it involved allocation of cables, buffers, and other resources in a path between two communicating entities, as indicated in the following figure.
|
Any of these resources could be destroyed in a war, breaking the connection, and requiring reconstruction of a path. Alternate paths might not be available, since circuit switching providers tend to concentrate the underlying physical circuits (cables, satellite links, etc.) in order to save money. So the ARPANET used a different kind of technology, called packet switching. Packet switching was the most basic technological contribution of the ARPANET. Packet switching is also the basis of the protocols used in the Internet. To understand how TCP/IP works, we must understand the basics of packet switching. |
|
Packet Switching In packet switching, data to be sent over a network is divided into many discrete chunks of data, each usually not more than a few thousand bytes long and each called a packet. Each packet is self-contained and holds all the information required to send it to its final destination. Each packet is routed from one computer to the next across the network until it reaches its final destination. Dedicated computers are normally used to route packets from place to place, much like a smart relay; each of these computers is called a router. Routers are connected to each other by a physical data carrier, such as copper cable, a fiber optic cable, or a microwave relay. This physical connection between two routers is called a link. Each networked computer that people actually use directly is called a host. A host is connected to the network in the same way as a router is, expect that a host usually has a single link to one network, while a router has several links to different networks. In the ARPANET, for example, each host was connected to a single local router, so the host needed to know only how to send data to the router; it was not involved in the packet routing decisions. A networked computer that is not used directly by people but that provides some sort of service, such as providing large amounts of disk space, is called a server. The same machine can be both a host and a server. A general term for a computer that may be a router, host, or server is node. Each network node has an address, which is a type of identifier, usually numeric. (IP addresses actually identify interfaces of links to nodes, not the nodes themselves, but most nodes have only one interface). A node that wants to send a packet to another node uses the address of the destination node to tell the intervening routers where the packet should go. As a packet travels through the network, it traces a path, or route, among the nodes of the network. A router decides which link to send a packet on next. Once data are divided into packets, the packets can be routed to their destinations in one of two ways: through virtual circuits or as datagrams. |
|
Virtual Circuits A virtual circuit is a connection that is established when one node begins to send packets to another node. All packets between the two nodes pass over the same route as long as that connection lasts, as indicated in the following figure. |
|
The figure shows two links selected from many in order to form a path between two nodes, E and B, going from E to A to B and back, with all traffic for the virtual circuit passing through that path in either direction. A node may be either routers or simply hosts or servers with more than one link. A protocol that implements a virtual circuit is called a connection-oriented protocol. Virtual circuit technology and connection-oriented protocols have been advocated by large segments of industry, academia, and government worldwide, especially by telephone companies. This is the kind of technology used in X.25 and other ISO (International Organization for Standardization) OSI (Open Systems Interconnection) protocols. |
Datagrams
However, a fast, efficient, dependable virtual circuit becomes none of those things when a node or link along its path fails, as either hardware circuit switching or virtual circuits will, particularly if failures are unpredictable in location, extent, or duration, as one would expect in a war. So ARPA chose to route packets individually, each as a datagram. A datagram is a packet with sufficient addressing information that it can be routed independently, as indicated in the following figure. The figure shows datagrams taking different paths at different times and in different directions. A protocol that uses datagrams without virtual circuits is called a connectionless protocol.
Intervening nodes do not usually allocate resources specific to datagrams traveling between a pair of nodes because any given packet might take a different route, thus bypassing a node that other packets traversed. This sounds inefficient, but it need not be. Packets can be rerouted for reasons other than node or link failures. For example, a link or node might become overloaded. A router then can choose to reroute some packets around the heavily loaded link. Use of datagrams thus permits load balancing throughout the network. This kind of dynamic reconfiguration is useful not only when nodes or links degrade or fail sue to enemy action, but also when this occurs as a result either of ordinary hardware or software overloads or failures, or of simple human error. The ARPANET was mostly built out of datagram technology because of its inherent fault tolerance.
|
Internet History The Internet evolved from the ARPANET, and sometimes people still confuse the ARPANET with the Internet. This confusion results probably because the ARPANET was the first backbone network of the early Internet and remained a part of the Internet until it was retired in 1990. But the Internet has always interconnected other networks and other kinds of networks. Some sites of interest |
|
NSFNET In 1984, National Science Foundation (NSF) established an office for networking and implemented several versions of its NSFNET national backbone network: in 1986 (DS-0, 56Kbps), 1988 (T-1, 1.544Mbps), and 1990 (T-3, 45Mbps). IN addition to NSFNET, NSF provided seed money for the NSFNET mid-level networks that are commonly known as the NSFNET regionals. These regional networks (now all mostly self-supporting) provide extensive connectivity for campus networks at educational institutions, government agencies, and commercial businesses. |