AIMD-Based Congestion Control for Layered Multicast

It is a key subject for layered multicast congestion control to promise fair bandwidth sharing. An AIMD-based congestion control algorithm for layered multicast is presented. The algorithm makes use of the AIMD algorithm, which is proved to have good TCP-compatibility and stability. To prevent rate oscillation by halving the sending rate in TCP, the algorithm adopts the principle of slow-increase and slow-decrease. We present a new feedback-free method for multicast receivers to estimate their round-trip time to the sender, which avoids complexity and scalability of feedback processing. A slow-start mechanism like TCP is adapted to promote bandwidth utilization and rate convergence. In addition, we give a congestion response strategy to solve the problems of how to response the congestion and how to re-allocate network resource. The strategy is relative to congestion signals and can provide good flexibility by selecting different parameters. The receivers will determine whether to subscribe the next layer or to unsubscribe the current layer according to their rates, which are updated using AIMD algorithm every a round-trip time. Our evaluation metrics focus fairness, bandwidth utilization and algorithm stability. The simulation results show that the algorithm achieves fair congestion control between TCP flows and multicast flows, and has high bandwidth utilization and good stability.