Threshold-based priority policies for parallel-server systems with affinity scheduling

Abstract

We study the dynamic scheduling of multiclass parallel-server stochastic networks and propose a general class of threshold-based priority policies. A number of examples are considered to illustrate the use of our algorithm for determining the placement of thresholds and to quantify the benefits of the proposed policies. We derive approximate formulas for the queue lengths under our threshold-based priority policies and we illustrate how these formulas can be used to obtain optimal threshold values.