Network Multicast Congestion Control Technology

With the speedy network technology and bandwidth improvement a number of new multimedia applications immerges, such as: video broadcast, video on demand, conference call and etc. for these new applications, traditional unicast and broadcast technology will result in big bandwidth waste. Multicast technology can significantly relieve the bandwidth consumption without changing current network structure and putting much investment. In the multicast congestion control field, multicast congestion control device has been given much attention.In this paper, we utilize a one-node to many-node multicast model and the data flow of campus network to design a congestion control device which is at a fixed sending rate and end driven.With the modern control theories: Schur-Cohn criterion and Routh-Hurwitz stability judgment method, we find the better parameters. The calculated results proves the device can make the source's transmitting rate steady and branches buffer occupancy at a stable volume.Then I established a multicast model in NS2, and add self-similar flow and exponential distribution flow. In the basis of calculated results by the device, we give different loads at source and study the branches average occupancy, max occupancy and real-time occupancy.With the simulation data, we could get the following conclusion: in self-similar flow model,when the load doesn't surpass 54.1%, there is no congestion in network, and the occupancy at a lower volume; in exponential distribution flow model,when the load doesn't surpass 92.7%, there is no congestion in network, and the occupancy at a higher volume. Self-similar flow is sharper than exponential distribution flow. This paper proves: network has a stronger capability to transport exponential distribution flow and the design of the device is reasonable. The results can give some advice when we set the occupancy volume and network load in campus network.