Measurement And Research Of Streaming Media Performance Based On End-System

Streaming media, as a combination of broad-band communication technology and multimedia technology, implements"Playing when Loading"for multimedia data through network, saving storage space and decreasing waiting time. As the packet-switch network is not designed for real-time, incessant streaming media data transmission, the violent changes of network bandwidth and jitter cause it can't satisfy with the real-time character of streaming media. The results makes the users sense badly because of playing stop, rebuffering and violent-shaking screen. So how to measure and improve the quality of service (QoS) of streaming media service effectively in Internet will get the important theory significance and application value.Firstly, this thesis analyses main reasons which influence the performances of streaming media service according to the application character of the streaming media technology. Chapter 3 proposes a set of metrics for streaming media service from application layer. The simulation results show that these metrics can express the sense of the users efficiently, and provide a reference for improving the steaming media service performance. The proposed measure method is simple, easy-deployed, independently, and doesn't depend on other subjective factors. It can reflect streaming media's quality more intuitively and veraciously.Secondly, the traditional data distribute form, which uses unicast and IP multicast, can not satisfy the requirements of new applications on bandwidth and QoS in large-scale streaming media system. From the aspects of End-System, the chapter 4 proposes a new application layer multicast algorithm based on the peer's performance evaluation - ALM based on Peer Performance Estimate (PPE) which derives from the idea of peer-to-peer. PPE organizes the multicast tree mainly by the peers'transmission ability and incorporates other related information as well. The algorithm makes the high performance peer close to the media source and the low performance one close to the leaf, which improves the efficiency of the multicast tree, and decreases the delay of peer-to-peer, link stress, link stretch and control overhead.Finally, the thesis introduces the RealMeasure, a tool used to obtain streaming media metrics from the application layer. RealMeasure is easy to deploy and provides the good character of expansion, openness, and necessary data for streaming media services.