Perturbation Analysis and Feedback Control of Communication Networks Using Stochastic Hybrid Models

Communication networks may be abstracted through Stochastic FluidModels (SFM) with the node dynamics described by switched flow equationsas various events take place, thus giving rise to hybrid automaton models withstochastic transitions. The inclusion of feedback mechanisms complicates thesedynamics. In a tandem setting, a typical feedback mechanism is the control of anode processing rate as a threshold-based function of the downstream node's bufferlevel. The problem considered here is to control the threshold parameters so as tooptimize performance metrics involving average workload and system throughputand to show how Infinitesimal Perturbation Analysis (IPA) can be used to analyzecongestion propagation through a network and develop gradient estimators of suchmetrics.