Circuit switching, packet switching

Circuit switching is where the connection across the network is a dedicated link that lasts for the whole transmission. This is how phone calls usually work: a phone is connected to a switching station, by a landline or by radio; that switching station is connected to other switching stations; and the last switching station is connected to the phone of the person you are calling.

In a circuit-switched system, all the data is sent from node to node across the same connections. So for the whole of the communication, those links cannot be used by other data.

Traffic on the internet is carried as packets. An internet packet is made of the data being carried, called the payload, and a header.

Internet packets carry all sorts of payloads: part of a web page, an email, or a streamed audio track, for example. The type of data is identified by the protocol field in the header. Internet packets have a maximum size, to stop anything "hogging" the bandwidth.

Packets are moved around the internet using a method called packet switching. Using fairly small packets lets many users share a data path. Different packets from the same "conversation" may be sent over different routes.

The internet follows an end-to-end principle: the end points (the source and the destination) are responsible for checking that everything that was sent has been received. This kind of communication between sender and receiver is called connectionless (rather than dedicated). Most internet traffic uses packet switching, and the internet is basically a connectionless network.

Questions often ask for differences between the two methods. These all come from the points above:

In any packet-switched network, the router is the key device. The router looks at the destination address of an incoming packet and decides which of its interfaces to send the packet to, until the packet finally arrives at its destination.

A home router (such as the one in your "home hub") has only two interfaces: the LAN and the WAN. Here the decision is simple: every packet that is not for the local network is sent to the WAN interface. Core backbone routers have many interfaces. Because the internet backbone is a mesh, it may be possible to reach a destination over several different paths.

Each router has a set of rules, a routing table, that it uses to decide what to do with an incoming packet. Each packet is treated on its own, and a separate routing decision is made for each one. Each router-to-router link is called a hop. The router works out the best "next hop" to move the packet on its way to its final destination. If any part of the communications infrastructure fails, the router chooses a different route, and later packets may be sent over a different path.

Routers can be set up to select the best route, or they can share data with each other so that they learn the best routes. This is how the core routers work. The routing table is constantly updated with information about the best route for packets to take.