Multi-objective routing in integrated services networks: A game theory approach

The multiobjective routing problem in multiple-class integrated services networks is presented. A multiserver two-class queuing model is introduced, where packets from the first class can be queued, while packets from the other class are blocked when the number of packets in the system exceeds some threshold. Therefore, the first class wants to minimize its average packet delay, while the other class wants to minimize its blocking probability. The resulting multiobjective routing problem is formulated as a Nash game, where each class tries to minimize its own cost function in competition with the other class. The routing policy for a two-server parallel system is derived and the strategy and performance of each class are shown.<<ETX>>