Master Thesis - 2012

Video distribution services via the Internet are widely provided due to improving network infrastructure and coding techniques. Among these services, live streaming service becomes popular in these days. For live video streaming, tree-based Application Level Multicasts (ALMs) are effective with respect to transmission fairness and jitter. In particular, the multiple-tree ALMs can alleviate excessive consumption of upload bandwidth among nodes. However, most conventional multiple-tree ALMs are constructed using a structured overlay. These ALMs unnecessarily increase transmission delay and consume substantial network resources because the structured overlay of live streaming system does not take into account distances in the IP network. In addition, the network constructed by a structured overlay has poor churn resilience because the network needs to reconstruct most of all distribution tree networks. Especially, when the massive nodes leave from the delivery network in short term, remaining nodes in the network suffer the stop of receiving the video data.

In this study, I propose the “overlapped cluster-trees” which is the method of the live streaming with churn resilience. The network constructed by the overlapped cluster-trees has the merits of the mesh-based and tree-based; churn resilience and short transmission delay. In the overlapped cluster-trees, one delivery network is divided into mesh-based clusters considering the distance of IP network at network construction. In addition, the clusters are organized into the multiple tree-based network. Moreover, I propose the optimum procedures for network construction in the overlapped cluster trees.

Finally, I evaluate about the proposed ALM by comparing with the conventional ALM through computer simulations. I show the effectiveness of the proposed overlapped cluster trees system. The results show that the proposed system can greatly enhance the churn resilience. In addition, the proposed system can decrease the transmission delay and resource consumption.