Low Bandwidth Consumption Research And Implementation Of P2p Live Streaming Systems

With the development of the computer and network technologies, streaming media has been fully applied since its advantage on playing while downloading. In recent years, P2P technology in streaming media has been getting mature, and P2P Live Streaming System had made tremendous achievements in scalability, robustness, and stability. But meanwhile plenty of P2P applications including P2P live streaming are consuming much precious bandwidth of bone network, and leads to the severe network congestion and some weaknesses like slow start-up. All these need to be improved. Therefore, in this thesis we presents a P2P live streaming system, Low Bandwidth-occupied Live system, LLive. The main character of this system is reflected on low bandwidth consumption in backbone, and it's composed of server, data sources, super nodes and user nodes.In the design of the LLive system, we first propose a coordinate marked system with hierarchy and dynamic organization, to achieve a matching relationship between physical and logical topology. We divide the partnership of nodes into neighbor partners and remote partners, so we could hold connection with outer-network nodes besides achieving topology matching. Then we introduced the notification/back-off mechanism in data scheduling, further reduce data transmission between networks and save bandwidth of backbone network. Apart from this, we design data resuming mechanism to accelerate the speed of data dissemination by pushing and pulling data. And then, the organizational structure of the LLive system is very flexible because it could support two modes, the enhanced mode which involve fixed super-node and the minimum model which has no fixed super-node. In the end we carry out a unified design in implementation of data source, supper nodes, and user nodes, and different types of nodes are generated after shielding specific functional modules, so the design is unified and reduce coding amount.We make experiments of LLive in both LAN and simulation environment. In the experiment of LAN, we verify the accuracy and availability of server, data source, super nodes and user nodes: LLive could play smoothly streaming media programs and has a shorter delay and startup latency; it could also effectively deal with the problem of packet loss, while the failure of random nodes have no significant effect on the broadcast. Through the experiment in the simulation environment, LLive system could quickly achieve the aggregation of nodes for the purpose of topology matching, and it could well adapt to the impact caused by the dynamically leaving of nodes in the marked system. After adopting the notification/back-off mechanism in data schedule, it could reduce more than 90% of the bandwidth occupancy in bone-network and achieve a more satisfactory result.