The Research Of P2P Streaming Transmission Mechanism In Heterogeneous Environment

Since quality of live streaming service has great relation with the streaming server's bandwidth, and the Client/Server (C/S) model has a potential performance bottleneck that a server with heavy load would lead to poor performance of the whole system, the C/S model cannot support streaming services with large numbers of users. P2P streaming system which distributes video data via P2P (Peer-to-Peer) network is an effective solution for such scenarios. Independing on network infrastructures, P2P streaming system can utilize upload bandwidth of user peers effectively, and then provide streaming service to a large number of users.Comparing to P2P file sharing system, P2P streaming system has more stringent real-time requirement for delivering data between peers, long transmission delay will lead to degradation of the streaming system's performance and quality of service. Although peers are equivalent in traditional P2P transmission model, Internet is a heterogeneous network. Access mediums of user nodes are different, and terminals of these nodes are various. In the viewpoint of protocol layers above network layer, features of such heterogeneous environments are presented as diverse access bandwidth and service capabilities of user peers. This thesis presents deeply studies in data transmission relationships and data scheduling strategies among peers, as well as transmission quality optimization mechanisms for P2P streaming over wireless multihop networks. The main contributions of this thesis are as follows:Since streaming data is copied and forwarded among peers via unicast tunnels in P2P streaming system, in order to reduce transmission delay and improve overall performance of the streaming system, data transmission relationships between peers should be optimized consciously. An effective P2P network topology is expected to be formed in application layer. This thesis proposes an adaptive topology optimization method based on bidirectional peer selection. Taking account of differences among peers'access bandwidth and service capability, the propsed topology optimization method improves overall performance of the streaming system by determining new peer's topology position according to its capability and contribution to the system. Meanwhile, the proposed method also considers benefits of user peers. In order to guarantee transmission quality between peers, each peer sends their service requirements to parent nodes which have minimum transmission delays. Experiment results show that the proposed method can guide peers constructing transmission relationships with high performance, and reduce the number of forwarding hops and transmission delay between the user peer and the source node.Because transmission delays among peers are various in heterogeneous environments, status of the user peers'download buffer will be asynchronous after numbers of forwarding hops in P2P network. The two major parts of viewing experience, playback delay and resilience, are related with the determination of data scheduling strategy between the user peer and all parent nodes, but they are two contradictory parts of the viewing experience, we cannot improve them at the same time. A viewing experience optimization method based on adaptive adjustment of requesting window is proposed to solve this problem. The proposed method utilizes heterogeneous real-time demands for different genres of videos, and then adjusts the requesting window adaptively according to the selected requesting strategy. For channels preferring to real-time experience, the proposal reduces playback delay under the condition of guaranteeing basic resilience; for channels preferring more resilience, resilience and activity of P2P data exchanging would be considered with higher priority. Experiment results show that comparing to the existing published strategies, the proposed method can achieve a reasonable balance between these two performance measures. Effectiveness of the proposal's adaptive feature is also proved in the experiment section.Existing research works on transmission quality optimization over wireless multihop networks usually fall into research of new routing algorithms. However, it's not realizable to require all nodes which consist of the wireless network executing the same specific routing algorithm. Moreover, P2P streaming system shouldn't have requirements to the underlying networks, as it's just an application. By analyzing the transmission model in wireless multihop networks, this thesis proposed a transmission quality optimization mechanism for P2P streaming system which works in the level of data topology and is independent on specific routing algorithms. Based on the features of H.264 video coding, the proposed mechanism adopts various priorities for different video slices when sending streaming data to its child nodes. Transmission qualities of video slices which are more important are guaranteed with higher priority, and growth ratio of control overheads is reduced by minimizing the number of UDP packets. Experiment results show that the proposal can improve the quality of decoded streaming data effectively. With the above as base, this thesis further proposes a goodput optimization framework which is independent on the realization of the P2P streaming system. In the proposed framework, in order to utilize spare bandwidth through a transmission path effectively, streaming requests generated by the naive streaming system are reassigned according to the qualities of each transmission paths connected with parent nodes. Experiment results show that the proposed framework can improve playback continuity of the streaming service significantly by introducing little control overheads.In the last chapter, a mesh-based P2P streaming system InfoTV which fits for heterogeneous and dynamic network environments is introduced in detail. The experiences in designing and implementing the system are valuable to this thesis and other P2P streaming researchers.