Research On The Performance Optimization Of P2P Transmission For Internet Massive Streaming

Internet is based on the point-to-point transmission and the problem of the point-to-multipoint transmission hasn’t been solved completely. To address the request forthe transmission performance by the streaming application, that is a typical point-to-multipoint transmission application, some scholars propose IP multicast, CDN and P2P.Because of the factors of cost and deployment, IP multicast and CDN are not suitablefor streaming transmission. With the good features, such as good scalability, low costand easy deployment, P2P has been the main technical approaches to solve the point-to-multipoint transmission problem of Internet massive streaming. However, the streamingapplications reveal a massive tendency, which present higher requests for the transmissionperformance of P2P and the current P2P transmission mechanism still has some deficien-cies in transmission performance, such as the throughput is not high enough, the trans-mission delay is not low enough and the inter-domain trafc are too much. To address theproblem of the transmission performance of current P2P, this thesis focuses on studyingthe performance optimization of point-to-multipoint transmission performance for mas-sive streaming based on P2P, and develops specific research from the following three mainfactors that afect the P2P transmission performance: topology formation, transmissionscheduling and trafc optimization. The main research contents and contributions of thisthesis are as follows:Firstly, we review the solutions of point-to-multipoint transmission in current Inter-net and hot research topics of current P2P streaming transmission. Through summarizingthe of current P2P transmission mechanism and exploring the main factors that afect theperformance of P2P streaming transmission, we propose a research framework of the per-formance optimization of P2P transmission for Internet massive streaming, which consistsof three aspects: transmission topology, transmission scheduling and trafc optimization.This research framework will guide this thesis to systematically carry out the research onthe optimization of P2P transmission performance.Secondly, we investigate the essence of overlay topology formation problem, simul-taneously considering the bandwidth and latency of peers. We propose a new neighborselection algorithm with biased QoS random walk, and to deal with the dynamic of P2P systems, we address an adaptive topology refinement algorithm. Through theoretical anal-ysis, compared with traditional method, our topology formation approach can efectivelyimprove the QoS performance of neighbors. The simulation results show that our approachcan obtain better neighbors and improve the transmission performance.Thirdly,we investigate the transmission scheduling algorithm for mesh-based P2Psystems. We address a pull and push transmission scheduling algorithm based on networkcoding and as for the heterogeneous environment of peer access bandwidth, we combinenetwork coding and SVC coding to design an inter layer joint coding scheme, and pro-pose a new transmission scheduling algorithm based ont this joint coding scheme. Thesimulation results demonstrate that our pull and push transmission scheduling algorithmbased on network coding can efectively improve the transmission performance and thetransmission scheduling algorithm based on joint coding scheme can efectively solve theproblem with the streaming transmission in the heterogeneous environment.Fourthly, we investigate the P2P inter-domain trafc optimization. As for live stream-ing, we propose a hybrid topology formation algorithm, combining the good property oftree and mesh, and design a transmission scheduling algorithm for this hybrid topology.As for VoD, we propose the distributed peer cache mechanism based on network codingand the cache replacement algorithm for proxy cache server. The simulation results showthat the inter-domain trafc of both the live streaming and VoD can be remarkably reduced,without decreasing the transmission performance.Finally, we design and realize a massive streaming P2P transmission prototype sys-tem, in which all the solutions proposed by this thesis have been included. Through thetest in the real network, the results demonstrate that our prototype system can significantlyimprove the transmission performance of traditional P2P streaming systems.