Flow-oriented multiple-session multicast routing on stochastic networks

Multicast is an efficient communication method to distribute high-bandwidth multimedia from a source to a group of receivers. The paper addresses the problem of routing multiple media streams to minimize the multicast tree cost and to satisfy all receivers' flow requirements in a capacity network having unreliable links and nodes. A multicast routing algorithm is proposed which is based on the idea of routing the source flows toward links with more bountiful bandwidth to cope with the topology and group membership changes due to the failures in the network. The algorithm finds a minimum-cost tree in O(n/sup 2/|D|) time, given an n-node network with |D| destinations. For evaluating the multicast algorithm on the network with unreliable nodes and links, a class of generalized random network, called stochastic network, is proposed. In comparison with modified versions of Dijkstra, Prim, and Esau-Williams algorithms, the simulation results show that our algorithm is the most robust in the sense that it has the highest probability of finding a feasible tree in the stochastic network. Also, it has the lowest average tree cost and highest throughput in case that the link connection probability is exponentially distributed, the nodal reliability is an exponential function of time, and the flow requirement is governed by the Poisson process.