Research Of Mesh-tree Double-layer Application Layer Multicast Model And Dynamic Degree-constrained Loading Balance Strategy

In recent years, under the background of network era, the internet technology is spreading rapidly which is motivated by both of the enterprise and user’s requirements. In the current network environment, neither traditional unicast nor IP multicast technology is able to satisfy the large amount of video data transmission and user’s high concurrent application. As a result, network television and video streaming system based on the application layer multicast are developed quickly and widely used.Application layer multicast technology not only avoids the low transfer efficiency of unicast, but also solves the problem of the IP multicast which is difficult to deploy. Besides, application layer multicast transfers the main task of multicast communication from router to local host. The local host plays the central role in the application layer multicast system. It makes the system more flexible, but its performance is uncertain. Its software and hardware performance, network environment and user characteristics can affect the overall performance of the multicast system directly. The key point of this thesis is to assess each end-host node’s comprehensive performance so as to balance the application layer multicast system’s load.Based on the research results both in china and abroad, this thesis makes a thorough analysis on the application layer multicast model and the loading balance technology. A mesh-tree hybrid double-layer multicast model network has been structured in the thesis. And a dynamic degree-constrained loading balance strategy (DDLBS) has been proposed based on the model. In mesh-tree hybrid double-layer multicast model, the multicast network is structured by hierarchical strategy. The model’s top layer is mesh model, under layer is tree model. During building the multicast network, some related factors will be considered comprehensively, such as end-host software and hardware performance, download and upload bandwidth, data transmission performance and user’s characteristic. And some evaluation algorithms are used for evaluating the end-host’s comprehensive performance. Then, DDLBS is carried out according to the different performance of each end-host node. By these processes, DDLBS adjusts the node’s degree and balances the node’s load dynamically, which reduces the whole load of multicast network and improves the stability of the multicast system. As a result, DDLBS adapts to the large amount of data transmission and high concurrency requirements well.