On equilibrium analysis of acyclic multiclass loss networks under admission control

We consider equilibrium analysis of several dynamic resource sharing policies for multiclass loss networks with acyclic topologies. The policies of interest are based on the principle of prioritizing classes via thresholding or reservation. We show that under each policy the equilibrium network state is a Markov random field and we obtain closed form expressions for the conditional probabilities therein. Such representations drastically reduce the computational complexity of blocking probability and revenue calculations. We provide revenue comparison of the considered policies and several extensions of the applied analytical technique.