Based On P2p Content Distribution System Of Streaming Media Services

The rapid development of the Internet infrastructure and the corresponding hardware technologies provide an important opportunity for the improvement of the Streaming Media Technology. However, as the current Internet bandwidth can not match the fast increasing requirement of user and the media files, the traditional C/S model can not provid satisfactory QoS. In this situation, P2P streaming media technology was developed. It decentralizes the service load to each node in the network by using the network resource of the clients. However, in a pure P2P network, since the high frequency of exit and join of nodes, as well as the complex topology, it is challenging to fully use the resource from clients and to build an efficient and stable streaming media service system with perfect play experience. On the other hand, Content Delivery Network (CDN) decentralizes the service load by replicating popular content on servers close to the clients, whereas the traditional CDN web servers still use C/S model to communicate with clients. This decreases the capacity of the system, because the capacity of those web servers is probably the bottleneck of the whole system. Therefore, the combination of both techniques, which means building a P2P network between web servers and the clients in CDN systems, can probably make fully use of their advantages.Based on in-depth research in both techniques, a Hybrid Streaming Media Service System, i.e. P2P-CDN system, which combined the P2P and CDN techniques to satisfy a high extendibility and a high quality of play experience, was proposed in this research. The system had two layers: the upper one was a backbone distribution overlay network, and the lower one was a P2P mesh used to connect the clients and the backbone network. In this case, the system could provide largescale live broadcasting and video on demand service. For the live broadcasting service, the play latency was minimized and the QoS was guaranteed through optimizing the overlay backbone's topology. As for the video on demand service, the QoS was increased using both push and pull strategies. Push strategy means that the overlay servers initiate the caching of popular file; while pull strategy represents that the clients pull the data from the servers. The system realized a better load balance under the load control of the Intelligent Publisher.Since a big synchronization lost latency exists during the live broadcasting in pure P2P streaming systems, a new Simulated Annealing Genetic Algorithm was proposed to solve the degree-constraint weighted minimum spanning tree problem. In this sense, a Simulate Anneal Arithmetic way was applied to stretch the fitness of chromosomes. Furthermore, more efficient crossover and mutation operators were used. This prevented the overhead of penalty and repair functions. As a result, the transmission latency on the backbone network was reduced. The simulation experiments proved that the algorithm created a shorter latency than the greedy algorithm did. And a good astringency was also obtained.For the load balance realized in the Intelligent Publisher subsystem, a simple but efficient algorithm, that is, dynamic forecast algorithm was proposed. This algorithm collected the load information of each overlay servers dynamically, combined them with the issue of file size, and finally divided the caching node. Using this algorithm, the load balance for video on demand service was realized, which was proved by the simulation experiments.