Perturbation analysis of multiclass multiobjective queueing systems with `quality-of-service´ guarantees

We consider a multiclass, multiobjective single-server queueing system where each class (associated with a “source”) has its own performance measure. The quality-of-service (QoS) guarantee problem requires the system to ensure that the performance objective of each class is always satisfied. Using perturbation analytic techniques for discrete event systems, we address the problems of predicting the effect of a source addition/removal on the performance objectives of all sources. We present two algorithms for solving these problems, the phantom source and marked source algorithms. Simulation results are included demonstrating the computational efficiency of this approach