Circuit switching is a technique for telecommunications networks in which information is transferred over a committed resource, such as a physical wire, radio frequency, or time slot in a fixed time-sharing schedule for transmission. Those resources are committed, whether used or not, for the duration of the association between source and destination.
The Public Switched Telephone Network is the classic example of circuit switching. Once a source and destination are associated by a telephone call, the resources for that call are committed even if neither participant speaks.
In contrast, packet switching methods conceptually do not commit resources to an association, but use shared resources only when there is information to transmit. Packet switching is the basic paradigm of the Internet.
Circuit switching had the great advantage of being feasible with electromechanical equipment, including manual interconnection point and electromagnetic switches, before the existence of electronic computers.
Circuit switching is compatible with physical sharing of resources within a transmission system that can contain multiple dedicated paths analogous to separate wires. For example, in frequency division multiplexing (FDM) over wired or wireless media, each connection is on a separate electromagnetic frequency. FDM works with both analog and digital signals.
In time division multiplexing, digital information is associated with a certain per-connection information transfer rate. The signal elements of each connection are assigned to a committed time slot within a physical medium that is has an aggregate transfer rate much faster than the per-connection rate. Such aggregates can be switched electromechanically or electronically, as in groups of telephone calls between switching offices.
There are newer methods that combine aspects of packet and circuit switching, such as Multi-Protocol Label Switching or paths set up with the Resource Reservation Protocol. Such methods do share media, perhaps with an automatic failover mechanism to change media in the result of failure, but endeavor to traffic for a particular source-destination pair on the same shared path through the duration of the association.
By doing so, the information in the path has constant latency imposed by speed-of-light and processing delays in the path. While these methods do not actually reserve resources, they usually are part of a admission control scheme that prevents assigning more traffic to a shared path than it can handle without significant queueing delays caused when packets cannot get to the path until other packets are sent. Queueing delays, in this context, refers to packets in different flows; it is perfectly appropriate to have systems that maintain packet ordering within the same flow.