Sensitivity Analysis for Discrete-Time Randomized Service Priority Queues∗

We consider a collection of queues operating in parallel and sharing a common server via a “randomized” service priority sharing rule. We assume that the service times of the customers (cells) are all constant and identical. This paper is concerned with estimating the sensitivity of the tail of the distribution of the typical sojourn time through these queues to the fraction of server bandwidth given to them by randomized service sharing. Two approaches are considered: “smoothed” perturbation analysis and stochastic intensity-based estimators.