Research On Low-Latency Topology In Peer To Peer Based Live Streaming Systems

In last few years, due to significant progress made on Peer-to-Peer network, P2P technology has been applied in many kinds of service, such as multimedia service based on streaming live media. As one of the most attractive research field, P2P live streaming is efficient on distribution of live streams in a real-time manner over a large population of spectators with good QoS and low cost, compared with traditional C/S model. Based on the research on recent P2P live streaming systems, which can be divided in terms of topology into two kinds: Tree-based and Mesh based, we propose a low-latency multicast protocol for each kind respectively.For Tree-based systems, we propose an application level multicast protocol for efficiently broadcasting live media streams to large-scale users. The main idea of this protocol is that end hosts are self-organized into a locality-aware overlay multicast tree based on their network position coordinates. It makes that the overlay multicast tree has a good match to the underlying physical network such that the link stress and the end-to-end delay can be effectively reduced, as well as handles some real-life network constraints such as limited bandwidth and connectivity constraint at the same time. In this paper, the rules for constructing and managing multicast tree are described in detail. Additionally, an experimental study is used to verify the advantages of the protocol. For Mesh-based systems, a locality-aware optimal Gossip protocol is proposed to construct a low-latency self-organized mesh. With a three-layer membership management scheme and the corresponding peer selection algorithms in each layer, every peer can fetch media data from the best partners in the network. Meanwhile, locality and proxy methods are employed in our proposal to traverse NAT, which enhance the system scalability. Finally, simulations are given to prove our protocol's efficiency.