Research On Application Level Multicast For P2P Streaming Media Service

Streaming media application is becoming increasingly popular on the Internet in recent years. The technological trend of the streaming media service system is that the scope of the resources to be managed is being enlarged to improve system scalability. For the P2P streaming media service system which integrates P2P technology and streaming media technology, the scope is being enlarged to all the dynamic, self-organized resources participated in the system and placed anywhere in the whole network. Under the well-designed resources management policies, P2P streaming media service system could satisfy the requirement of large amounts of users.How to solve the resource acquiring problem is critical to the P2P streaming media service system, and application level multicast technology is a simple and efficient resolution. However, due to the characteristics of the streaming media service and P2P system, how to construct more scalable application level multicast becomes a great challenge.Key to the application level multicast constructing is to solve three basic problems, which are design and implement of the problem, member management and self-adaption.In order to design the problem well and implement it in a centralized manner, we design a MSVMT problem model to aim at the delay- and degree-bounded constraint condition of the live video streaming application. We prove that the problem is NP-Complete, and prove its equality to the well-studied MMDMT problem under some condition. Then we propose a greedy algorithm FindALMTree and the PPAF heuristic evaluation function it adopts, where the PPAF heuristic could balance the delay and out-degree topology information integratively. We analyze amd prove the approximation degree of the greedy algorithm.In order to solve the member management problem and implement it in a distributed manner, we first design a tree-first application level multicast protocol called OverStream. When constructing the data delivery topology, it adopts the PPAF heuristic evaluation function to optimize the choice of the father node on the multicast tree. Its node join process could avoid load unbalance problem effectively, and its node self-optimizing process could guarantee the optimization of the tree structure and the continuity of the playing process. Simulation results show that the load balance and scalability of the OverStream protocol are better than the other congeneric protocols. Base on OverStream protocol, we design another application level multicast protocol named as LiveStream. LiveStream protocol adds a backbone topology which is lying between the control topology and data delivery topology. By the backbone topology construction technique based on the triangle rule, we can construct a topology-aware overlay network based on the physical network which can reduce the transmission delay of streaming media.In order to solve the self-adaption problem, we propose a new method to construct application level multicast based on virtual node, and we also propose a construction algorithm to construct an application level multicast tree based on anti-n-cube. We prove that making anti-n-cube as virtual node has better reliablility, scalability and internel delay consistency. Then we could sovle the play halt or play jitter problem due to the strustructure instability of the multicast tree.Finally, we design and implement the MagicStream streaming media service system. MagicStream system not only adopts application level multicast protocol, but also adopts multi-source collaborating streaming schedule technology. With index information and segment striping, MagicStream system could solve the VCR problem effectively. With segment local cache policy and segment striping, MagicStream system could solve the resource placement problem under the constraint that the resources of the nodes are bounded.