Packet versus Circuit Switching

TCP/IP networks route data between hosts in packets and packets and packets from many hosts are interspersed on a link, This is called packet switching.

Packet switching can be contrasted to circuit switching in which a circuit is exclusively dedicated to a session for its duration. The telephone network is a familiar example of circuit switching. A dedicated physical circuit is established between two points from the time a call begins until it ends.

Packet switching is more efficient because the cost of the link is shared by many users. When you make a circuit-switched phone call, the line is tied up for the duration of the call whether you are talking or silent or on hold.

The advantage to circuit switching is that performance is predictable and can be guaranteed. A flood of packets from another application cannot temporarily degrade performance. The quality of service (QOS) is predictable.

In asynchronous like e-mail or file transfer, such fluctuations are probably not of great concern, but for isochronous applications like voice over IP (VOIP), they can cause perceivable delays and lost or garbled words. Delays are also bothersome in interactive applications where a user is waiting for a reply from a server.

Circuit switching has traditionally been used for telephone communication, but the trend is toward packet switching. Generally speaking, call quality is lower on the Internet than on the public, switched telephone network (PSTN), but they may be indistinguishable on an intranet, and VOIP systems generally offer more features and better integration between data and voice applications. VOIP servers now outsell conventional telephone switches for use within organizations.

Finally, note that many hosts share the same link because they send their packets at slightly different times. The sharing of a link is called multiplexing, and this is an example of time-division multiplexing.