This chapter discusses telecommunication and information systems. However, there often remains the problem of characterising the user inputs in sufficiently simple terms to enable the analysis to proceed. As we have seen, this is not always a simple process. Then, assuming the user inputs have been dealt with, the main problem facing the system designer is not so much the lack of relevant mathematical theory but how to apply the theory to the system being designed. Queuing theory has been developed extensively and a considerable amount of information is now available to designers.

There are a number of approaches that can be used to obtain performance measures from systems and networks. One is to actually measure the performance of the system, but, unfortunately, one cannot ask 20 million telephone subscribers in the UK not to use their telephones one (or more?) day a year to allow experiments to be carried out! Another possibility is to use mathematical analysis, either by hand or on computer, to determine the performance measures of interest. The third is to simulate the system using a computer model. In this chapter, simulation will refer to the use of a digital computer program that attempts to duplicate the 'real' system. The computer simulation is driven with random inputs supplied by a (pseudo) random-number generator; and the results produced by the simulation are thus stochastic in nature.

This chapter discusses the queuing model. A queuing network model specifies the exogenous arrival process entering fresh jobs at each node, the routing rule which determines the next node to be visited by a job which has just completed service in a node, the structure of the service facility and the queue discipline in each node, and the successive service demands made by a job at each node it visits. It may also be useful to divide jobs into different classes, each with its own routing rule and different service characteristics at each node, and even allow jobs to change class during their stay in the network.

Message switched systems have been known since the early days of telegraphy. The simplest system was one in which telegrams or messages were received over a link in the form of punched tapes in the Morse or teleprinter code. The tape could then be used in a transmitter which sent the message over the next link towards its destination. In message switching, each switching node stores the incoming message and then forwards it over the outgoing route. Long messages in such a store-and forward switching strategy can impose large queuing delay on other users sharing the route. Thus, in contrast to circuit switching, where in the event of heavy traffic, calls are blocked, in message switching calls face long delays. To reduce delay, a message can be split into several short notes, called packets. In this case, before a complete message from one source has been fully routed, other users can interleave parts of their messages in packet forms. Hence, with packet switching the queuing delay is more fairly distributed among the users sharing the switching node. Packet switching can be connectionless or connection-oriented. In connectionless switching, like message switching, packets are stored and forwarded. This technique is also called datagram. The outgoing route can change from time to time, depending on the routing strategy. Hence packets may arrive out of sequence. Reordering of the packets to be assembled into the original message may cause additional delay. In contrast, in connection-oriented switching, all packets belonging to a message always follow the same route, so avoiding the out of sequence problem. In this case, prior to data transfer, a link between the source and destination is set up. This is similar to the call set up phase in circuit switching. Also to minimise overhead information, only an abbreviated 'address' need be given to obtain the full address and routing instructions, which are stored at each node when the connection is set up.

Although performance engineering originated in the study of telephone calls randomly originated by subscribers, the applications have been greatly extended in recent years. There are many applications outside the field of information engineering such as the study of road traffic, particularly the problems of queuing at roundabouts and obstructions. These are best dealt with in books and papers pertaining to the applications. Even within the field of information engineering, it would be impossible to cover the complete range of applications of performance engineering in a book of this size. Chapter 8 has covered the important field of reliability and this chapter gives a few more examples to illustrate the techniques.