The Research Of Minimum Delay Application Layer Multicast Tree Construction Based On Stability Probability And Loss Recovery Algorithm

As the increase of internet popularity and the surge of the scale of users, the Internet has infiltrated our life and work silently and soundlessly, it has changed and improved the mode of our life and work in many aspects. The traditional unicast needs to establish a unique data channel between the server and clients, that is to say the server and the clients establish the "one to one" communication mode, this mode can't satisfy the demand of data transfer efficiency of multimedia data. The appearance of IP multicast has resolved the problem of low efficiency of unicast communication mode. But IP multicast has changed the design principles of Internet that based on unicast, and need to change the existing network facilities, these problems have led the slow deployment of IP multicast all over the world. Application layer multicast doesn't need the support of special facilities, and ALM (Application Layer Multicast) has high flexibility compare to IP multicast, thus ALM will develop with bright prospect.ALM rely on the end hosts to copy and transfer multimedia data which have high dynamics, due to its constituent elements are different from IP multicast, the performance and efficiency of ALM are inferior to IP multicast, and ALM has many other aspects of problems. How to construct ALM tree with high stability and lower delay has always been a hot research topic in multimedia field.This paper mainly research on two aspects of ALM, one is the construction of high stability and low delay ALM tree, the other one is the high-performance failure recovery algorithm for ALM. The main research work includes the flowing:First, due to the transfer efficiency and stability of ALM are inferior to IP multicast, thus the construction of high stability and low delay ALM tree is the key to the spread of ALM. Aiming at this problem, this paper has proposed the SDMD (Spanning tree based on stability probability, degree-constrained, and minimum diameter for ALM) problem model, in this model we have redefined the definition of node's stability probability, and prove SDMD to be NP-hard. Therefore, we propose a heuristic algorithm to solve this problem. Our simulations demonstrate that our proposed algorithms can reduce the average receiving delay and cut down accumulative interrupt times effectively.Second, the high efficiency loss recovery algorithm is import to ALM tree when there happens a node leaves the multicast tree. Aiming at this problem, this paper proposed a hybrid partition based method for loss recovery (HPLR), in this method, we redefine the server ability of node, and divide the tree into two parts according to node's server ability, and each part of the tree has its own recovery method respectively, so as to achieve a balance between computational overhead and time overhead. The simulations demonstrate that the HPLR algorithm can effectively reduce the rejoin time after the nodes leave.