PDF of this tutorialIn principle, two basic technologies are used for building high-capacity networks: circuit switching and packet switching. In circuit-switched networks, network resources are reserved all the way from sender to receiver before the start of the transfer, thereby creating a circuit. The resources are dedicated to the circuit during the whole transfer. Control signaling and payload data transfers are separated in circuit-switched networks. Processing of control information and control signaling such as routing is performed mainly at circuit setup and termination. Consequently, the transfer of payload data within the circuit does not contain any overhead in the form of headers or the like. Traditional voice telephone service is an example of circuit switching.
Circuit-Switched Networks
An advantage of circuit-switched networks is that they allow for large amounts of data to be transferred with guaranteed transmission capacity, thus providing support for real-time traffic. A disadvantage of circuit switching, however, is that if connections are short-lived—when transferring short messages, for example—the setup delay may represent a large part of the total connection time, thus reducing the network's capacity. Moreover, reserved resources cannot be used by any other users even if the circuit is inactive, which may further reduce link utilization.
Packet-Switched Networks
Packet switching was developed to cope more effectively with the data-transmission limitations of the circuit-switched networks during bursts of random traffic. In packet switching, a data stream is divided into standardized packets. Each contains address, size, sequence, and error-checking information, in addition to the payload data. The packets are then sent through the network, where specific packet switches or routers sort and direct each single packet.
Packet-switched networks are based either on connectionless or connection-oriented technology. In connectionless technology, such as IP, packets are treated independently of each other inside the network, because complete information concerning the packet destination is contained in each packet. This means that packet order is not always preserved, because packets destined for the same receiver may take different paths through the network. In connection-oriented technology such as asynchronous transfer mode (ATM), a path through the network—often referred to as a logical channel or virtual circuit—is established when data transfer begins. Each packet header then contains a channel identifier that is used at the nodes to guide each packet to the correct destination.
Figure 1. Technology Overview
In many aspects, a packet-switched network is a network of queues. Each network node contains queues where incoming packets are queued before they are sent out on an outgoing link. If the rate at which packets arrive at a switch point exceeds the rate at which packets can be transmitted, the queues grow. This happens, for example, if packets from several incoming links have the same destination link. The queuing causes delay, and if the queues overflow, packets are lost, which is called congestion. Loss of data generally causes retransmissions that may either add to the congestion or result in less-effective utilization of the network. The ability to support real-time traffic in packet-switched networks thus calls for advanced control mechanisms for buffer handling and direction. As a result, the complexity and necessary ability to process information, and therefore the need for computer power, increases sharply when striving for high transmission capacity.
