What is performance, and why is it that performance modelling is different in principle from other types of modelling work?
In almost all other areas of telecommunications, modelling is devoted to determining the equilibrium mean behaviour of a system. Random fluctuations and perturbations are distracting factors which possibly must be allowed for, but that is done by robustness studies, or a series of planned variations to see the impact that they have around some fixed point. If possible, it is preferable to eliminate their effects at the outset. For performance, by contrast, it is the randomness that is the core feature to be studied: the entire area is devoted to examining the quantitative effect that randomness of input or environment has upon a system or network.
This means that performance models need a degree of accuracy which is unique among modelling techniques. Not only must they represent with considerable accuracy the target system being studied, but, in order to quantify what are essentially second-order effects, they must also take one of two courses:
either these effects must be directly (i.e. analytically) calculable from some more robust base-level quantities;
or the model must be so faithful that significant differences in output values can be confidently ascribed to the effects we are studying, and not instead to mere model inaccuracy.
The first choice above leads to the great category of analytic modelling techniques, of which the primal example is that of representing...
TABLE OF CONTENTS 
