Research Of The Congestion Control Based On Layered Multicast

Multicast is a new network application in Internet, it can save much of the bandwidth,so it is used to transmit Multimedia Data, such as video conference and remote-teaching.Layered multicast is an efficient approach in multicast communication. However the conventional layered multicast is cumulative, in order to decode the original data stream, the receiver should join the layers gradually,and it's rate can only be the sum of group's rate which it receives , and can not realize fine-grained multicast,so can not utilize the bandwidth sufficiently. Based on the research of the cumulative layered multicast, a multicast policy using non-cumulative layering is proposed, the original data is divided into several "equal" multicast layers, and the receiver can join a set of layers due to its bandwidth, but not to join the layers gradually from low layers to high layers, this can make the receiver join or leave the layer with more flexibility, so the receiver can utilize the bandwidth with great sufficiency. The multicast policy using non-cumulative layering is realized by MDC (Multiple Description Coding), we use exponential sequence to allot the rate in the non-cumulative layers, so it can provide the largest range by using certain layers.A receiver-driven TCP-Friendly congestion arithmetic has been proposed as well. At first, the arithmetic expand the RTCP protocol (Real-time Transport Control Protocol), and it can make the receiver get the Round-Trip Time of link between the sender and itself, then calculate the TCP-Friendly receive rate, and adjust the layers it should subscribe according to receive-rate of itself. Thus, the receiver can adjust the rate slightly and make the better of the bandwidth. The arithmetic was being emulated in NS-2 at last, the result show that the arithmetic can make the better of the bandwidth and scalable.