On Reducing Delays in P2P Live Streaming Systems

(ABSTRACT) Peer-to-peer (P2P) technology provides a scalable solution in multimedia streaming. Many streaming applications, such as IPTV and video conferencing, have rigorous constraints on end-to-end delays. Obtaining assurances on meeting those delay constraints in dynamic and heterogenous network environments is a challenge. In this proposal, we devise a streaming scheme which minimizes the maximum end-to-end streaming delay for a mesh-based overlay network paradigm. We first formulate the minimum-delay P2P streaming problem, called the MDPS problem, and prove its NP-completeness. We then present a polynomial-time approximation algorithm to this problem, and show that the performance of our algorithm is bounded by a ratio of O(√ log n). For a practical deployment, we extend the algorithm to a distributed version with the adaptation to network dynamics. Our simulation study reveals the effectiveness of our algorithm, and shows a reasonable message overhead. Besides playback lag, delays occurring in P2P streaming may arise from two other factors: node churn and channel switching. Considering the fact they both stem from the reconnecting request in churn, we call them churn-induced delay. A typical channel switching delay costs around 10 seconds or even more. Current users have been accustomed to delays under seconds, which are typical in a cable TV system. Thus, this long switching delay has negatively affected the extensive commercial deployment of P2P systems. In this proposal, we propose an agent-based scheme to provide preventive connections to all channels. Once an actual connection is requested, time will be saved in retrieving bootstrapping information and obtaining authorization as well as authentication in new channel. Towards an efficient control and message overhead, each channel will select some powerful peers as agents to represent the peers in the channel. Agents will distill the bootstrapping peers with superior bandwidth and lifetime expectation to quickly serve the viewer in the initial period of streaming. We build a queueing theory model to analyze the agent-based scheme.