Solving QoS multicast routing with genetic algorithms

Proliferation of group-based real-time applications, such as online games and video conferencing motivates research into QoS multicast routing. This type of applications requires consideration of both end-to-end delay (i.e., packet delay from the source to all destinations is bounded) and group synchronisation (i.e., the difference in packet delay from the source to different destinations is bounded) constraints. In this paper, we describe the combined problem of multicast routing and delay partitioning with end-to-end delay and group synchronization constraints in a QoS framework where a delay dependent cost function is associated with each network link [H.T. Tran and R. J. Harris, 2003]. Due to NP-completeness of this problem, a genetic algorithm (GA) based algorithm, that computes a source-based multicast tree that meet both requirements with near-optimal cost, is developed. In our GA, we compare two different tree encoding techniques: link weight and link bias encoding, and by the means of simulation, find that link weight encoding is less effective than link bias coding. The simulation result also shows that our GA consistently performs better than two other simple heuristics.