Study Of Application Layer Multicast Based On Available Bandwidth Measurement

Multicast is a communication way of one point to many points on Internet. Having been compared with unicast, multicast technology could save the bandwidth, lighten the network burden and improve the data transmission rate. As the next generation Internet is coming, Multicast technology plays an important role in audio-stream and video-stream application. However, traditional IP multicast hasn't been popularized due to the reason of the market and technology. Researchers begin to consider utilizing network resources of Internet end user to provide multicast service directly in the application level. It has no use for altering the network configuration. The packets are forwarded by end system not the routers.First, the thesis introduces the basic idea of Application Layer Multicast, which compares and analyzes the performance of main ALM protocols. Current ALM scheme has constructed an overlay among nodes on the application layer. While sending the data, the overlay completely shields actual physical topology on the network level. Its performance weakness is summarized. The thesis brings forward an available bandwidth measurement method based on multicast resource. Depending on the relation between the input data rate and the one-way delay, it uses multicast data itself (e.g. video code stream) as the probing source, and calculates the available bandwidth. The algorithm provides the support of high bandwidth, route efficiency, and low delay multicast services.Second, on how to construct a delivery tree, the available bandwidth measurement and the construction of multicast delivery tree is combined. The thesis provides a new ALM protocol based on the topology state awareness. Its core thought is to construct the multicast spanning tree through the measurement of actual link available bandwidth and latency. Having compared with the performance of Narada spanning tree, the simulation demonstrates the effectiveness of the proposed algorithm in terms of throughput and average link stress.Finally, an ALM group management including new nodes entry, exit and tree optimization is studied.