Performance bound hierarchies for queueing networks

In applications of queueing network models to computer system performance prediction, the computational effort required to obtain an exact equilibrium solution of a model may not be justified by the accuracy actually required.In these cases, there is a need for approximation or bounding techniques that can provide the necessary information with less computational effort.This paper presents a new technique that yields performance bounds for single-class separable queueing networks consisting of fixed-rate and delay service centers.Unlike previous approximation or bounding techniques, there is a smooth trade-off between computational effort and accuracy.Any level of accuracy (including the exact solution) can be guaranteed by investing the necessary computational effort.Performance bounds that are sufficiently tight for most practical purposes may be obtained with a fraction of the effort required for the exact solution.Since bounds are produced, as opposed to approximations, guarantees about the accuracy of a model solution can be provided.