An infinitesimal perturbation analysis algorithm for a multiclass G/G/1 queue (Q916216): Difference between revisions

Simulation of complex systems such as queueing networks is often the only approach for obtaining information about system characteristics. When it is desired to perform sensitivity analysis with respect to changes in n parameters, say, this can prove costly as it involves \(n+1\) simulations. Perturbation analysis is a technique for performing sensitivity analyses by perturbing the sample paths of one simulation. Problems with the technique arise when such perturbations lead to a change in the order of events, such as a change in the order of arrival of customers. A revised form of the algorithm for a multiclass GI/G/1 queue which copes with such problems is proposed and its properties are analysed theoretically and experimentally.