On QoS multicasting performance in wide area networks

Multicasting enables applications to scale to a large number of users without overloading the network and server resources. With the advent of multimedia applications, the focus of multicasting research has shifted from minimizing the overall cost of the multicast tree to finding one which supports the QoS requirements of the underlying multimedia application. Finding such a tree, however, is NP-complete. Several heuristics, such SPH, KPP, BSMA, and K-SLIM, have been proposed as an approximation of the optimal solution to the multimedia multicasting problem. These heuristics differ in their complexity, overhead and the way they minimize tree cost and end-to-end delay. This paper develops a simulation framework to study and compare the performance of these heuristics. Using the above framework, the multimedia multicast heuristics were tested with respect to the graph size, the multicast group size, and the delay requirements of the underlying multimedia traffic. The simulation results show that, on average, K-SLIM outperforms the other simulated heuristics. Furthermore, the results also show that the average cost of the multicast trees produced by SLIM+, a variation of K-SLIM which requires must less overhead, is close to the average cost of the multicast trees produced by K-SLIM.