Master Thesis - 2007

Recently, a lot of applications to deliver a high-quality video and voice data to many clients are used with capacity increase of a network. It is thought that the applicationlevel multicast that uses multi distribution trees is effective in respect of the low delay and efficiency. However, if the node failure and departure happens, as well as one tree-based, multi tree-based method does not have effective means to maintain the quality of the other nodes.

There are two methods to reduce the influence of the node departure, the one is the reactive method, and the other is the proactive method. The reactive method means that the children node prepares the alternative parent node in anticipation of the current parent node is troubled. If the failure occur, the children node receive the data from that alternative parent node. The proactive method means that the every node forward the extra data through the path which is different from the regular distribution tree. If the parent node is troubled, this node may receive data from other node. So, we think that the proactive method is more suitable for the real-time video streaming because that method does not need the time to detect the failure of parent node and the process to prepare the alternative-node.

In this paper, we adopt the random forwarding which base on the proactive method to the multi tree-based method. So, each distribution tree which operates independently in the existing schemes cooperate. We intend that this scheme enable to reduce the influence of the node departure and maintain the high connectivity of the whole distribution trees while maintaining a low delay and ease of the management. Moreover we evaluate the performance of proposal methods through the computer simulation. In addition, we evaluate the latency and overhead data which are become higher as the trade-off of the high connectivity, and clear that it is acceptable for the general applications.