electronic system of links and switches, and the controls that govern their operation, that allows for data transfer and exchange among multiple users.

A switched communications network transfers data from source to destination through a series of network nodes. switching can be done in one of two ways. In a circuit-switched network, a dedicated physical path is established through the network and is held for as long as communication is necessary. An example of this type of network is the traditional (analog) telephone system. A packet-switched network, on the other hand, routes digital data in small pieces called packets, each of which proceeds independently through the network. In a process called store-and-forward, each packet is temporarily stored at each intermediate node, then forwarded when the next link becomes available. In a connection-oriented transmission scheme, each packet takes the same route through the network, and thus all packets usually arrive at the destination in the order in which they were sent. Conversely, each packet may take a different path through the network in a connectionless or datagram scheme. Since datagrams may not arrive at the destination in the order in which they were sent, they are numbered so that they can be properly reassembled. The latter is the method that is used for transmitting data through the Internet.

Since all nodes can hear each transmission in a broadcast network, a procedure must be established for allocating a communications channel to the node or nodes that have packets to transmit and at the same time preventing destructive interference from collisions (simultaneous transmissions). This type of communication, called multiple access, can be established either by scheduling (a technique in which nodes take turns transmitting in an orderly fashion) or by random access to the channel.

In a scheduling method known as time-division multiple access (TDMA), a time slot is assigned in turn to each node, which uses the slot if it has something to transmit. If some nodes are much busier than others, then TDMA can be inefficient, since no data are passed during time slots allocated to silent nodes. In this case a reservation system may be implemented, in which there are fewer time slots than nodes and a node reserves a slot only when it is needed for transmission.

A variation of TDMA is the process of polling, in which a central controller asks each node in turn if it requires channel access, and a node transmits a packet or message only in response to its poll. “Smart” controllers can respond dynamically to nodes that suddenly become very busy by polling them more often for transmissions. A decentralized form of polling is called token passing. In this system a special “token” packet is passed from node to node. Only the node with the token is authorized to transmit; all others are listeners.

Scheduled access schemes have several disadvantages, including the large overhead required for the reservation, polling, and token passing processes and the possibility of long idle periods when only a few nodes are transmitting. This can lead to extensive delays in routing information, especially when heavy traffic occurs in different parts of the network at different times—a characteristic of many practical communications networks. Random-access algorithms were designed specifically to give nodes with something to transmit quicker access to the channel. Although the channel is vulnerable to packet collisions under random access, various procedures have been developed to reduce this probability.

One random-access method that reduces the chance of collisions is called carrier sense multiple access (CSMA). In this method a node listens to the channel first and delays transmitting when it senses that the channel is busy. Because of delays in channel propagation and node processing, it is possible that a node will erroneously sense a busy channel to be idle and will cause a collision if it transmits. In CSMA, however, the transmitting nodes will recognize that a collision has occurred: the respective destinations will not acknowledge receipt of a valid packet. Each node then waits a random time before sending again (hopefully preventing a second collision). This method is commonly employed in packet networks with radio links, such as the system used by amateur radio operators.

It is important to minimize the time that a communications channel spends in a collision state, since this effectively shuts down the channel. If a node can simultaneously transmit and receive (usually possible on wire and fibre-optic links but not on radio links), then it can stop sending immediately upon detecting the beginning of a collision, thus moving the channel out of the collision state as soon as possible. This process is called carrier sense multiple access with collision detection (CSMA/CD), a feature of the popular wired Ethernet. (For more information on Ethernet, see computer: Local area networks (computer).)

Since collisions are so detrimental to network performance, methods have been developed to allow multiple transmissions on a broadcast network without necessarily causing mutual packet destruction. One of the most successful is called spread-spectrum multiple access (SSMA). In SSMA simultaneous transmissions will cause only a slight increase in bit error probability for each user if the channel is not too heavily loaded. Error-free packets can be obtained by using an appropriate control code. Disadvantages of SSMA include wider signal bandwidth and greater equipment cost and complexity compared with conventional CSMA.

The application layer is difficult to generalize, since its content is specific to each user. For example, distributed databases (database) used in the banking and airline industries require several access and security issues to be solved at this level. Network transparency (making the physical distribution of resources irrelevant to the human user) also is handled at this level. The presentation layer, on the other hand, performs functions that are requested sufficiently often that a general solution is warranted. These functions are often placed in a software library that is accessible by several users running different applications. Examples are text conversion, data compression, and data encryption.

User interface with the network is performed by the session layer, which handles the process of connecting to another computer, verifying user authenticity, and establishing a reliable communication process. This layer also ensures that files which can be altered by several network users are kept in order. Data from the session layer are accepted by the transport layer, which separates the data stream into smaller units, if necessary, and ensures that all arrive correctly at the destination. If fast throughput is needed, the transport layer may establish several simultaneous paths in the network and send different parts of the data over each path. Conversely, if low cost is a requirement, then the layer may time-multiplex several users' data over one path through the network. Flow control is also regulated at this level, ensuring that data from a fast source will not overrun a slow destination.

The network layer breaks data into packets and determines how the packets are routed within the network, which nodes (if any) will check packets for errors along the route, and whether congestion control is needed in a heavily loaded network. The data-link layer transforms a raw communications channel into a line that appears essentially free of transmission errors to the network layer. This is done by breaking data up into data frames, transmitting them sequentially, and processing acknowledgment frames sent back to the source by the destination. This layer also establishes frame boundaries and implements recovery procedures from lost, damaged, or duplicated frames. The physical layer is the transmission medium itself, along with various electric and mechanical specifications.