Realization probabilities. The dynamics of queuing systems (Q1313415)

From the author's introduction (Chapter 1): ``The conventional approach used in the queueing system theory to obtain the steady-state performance sensitivity is mainly based on the probability and the stochastic process theories. With this approach, the explicit form for steady-state probability is first derived; the formula for the steady-state performance is then obtained; and the performance sensitivity with respect to a system parameter is finally obtained by taking the derivative of the performance formula. This book presents another approach to derive the steady-state performance sensitivity formulas. The approach is an analogue to the linearization of a perturbed nonlinear continuous variable system along its trajectories; it is based on the dynamic nature of queueing systems. The main idea of this approach is to study the final effect of a small perturbation in a sample path of a queueing system on a performance function. This final effect is measured by a quantity defined as the realization probability or its extension, the realization factor... A set of linear equations can be derived for realization probabilities, and performance sensitivity formulas can be obtained in terms of these realization probabilities... The approach explores the dynamic feature of a queueing network... the results obtained by this approach provide a new perspective of queueing systems... First, since the realization probability measures the final effect of a small perturbation on a performance function, it contains more information than the parametric sensitivity... Next, the new approach is based on studying the evolution of sample path of a queueing system... This feature is very important because for most systems there is no closed form solution for the steady-state performance, and simulation is the only way to obtain the performance sensitivity... Finally, based on this approach, explicit forms can be obtained for the sensitivity in some networks... As an application of the realization probability, we shall show that these sensitivity formulas lead to some important optimal control policies. The book applies the dynamic point of view to the study of queueing networks and can be viewed as a complement to the existing queueing theory books.'' The book is organized as follows. Chapter 2 ``Mathematical background'': review of the fundamental mathematics that are related to the topics presented in the book; Chapter 3 ``Realization probabilities in closed Jackson networks'': the dynamic approach is applied to study closed Jackson networks; Chapters 4-6 ``Realization factors in state-dependent networks'', ``Systems with general distributions'', ``Multiclass networks and networks with blocking'': the concepts and results of Chapter 3 are extended to other networks and to general performance functions: In Chapter 7 (``Sensitivity analysis of generalized semi-Markov processes'') the results are extended further to the generalized semi-Markov processes. In Chapter 8 (``Operational sensitivity analysis'') so-called ``operational'' approach is used to derive the sensitivity formulas by using ``operational variables'' such as the ``realization ratio''. Chapter 9 contains the application of realization theory to the optimization problems in queueing theory.