Overview

The previous chapters have given prominence to the power and efficiency of the analytical approach to performance evaluation. However, many realizations make use of such specific mechanisms so that their analysis is beyond the scope of mathematical tools. In that case, the simulation method provides an efficient way to overcome the difficulty.

A simulation experiment aims at reproducing the dynamic behaviour of the system (customers, servers, etc) through a computer program, most often run on specific software tools. On this software model, observations are performed, which give the figures of interest, such as mean or variance of delays, loss probabilities, etc. Clearly, the goal is not to reproduce exactly the microscopic level of detail of the original system, but to take account of those peculiarities in the mechanisms responsible of the overall behaviour. It takes a careful analysis of the system, and a good understanding of the queueing phenomena to extract the strict set of details compatible with the required degree of precision and the budget set for the study, expressed both in terms of the time needed to elaborate and validate the model as well as the duration of each simulation run.

Actually, simulation happens to offer quite a powerful approach for the engineer: it not only allows a quantitative analysis of complex systems to be made, but it provides a software prototype of the system under development, whose role is the same as the experiment for the researcher. Indeed thanks to simulation one can observe and...