ReDePoly: reducing delays in multi-channel P2P live streaming systems using distributed intelligence

Video streaming over peer-to-peer (P2P) networks is a promising approach for scalable IPTV and online games. Distributing high-quality videos among users who employ numerous channels in multi-channel P2P live streaming systems, however, still suffers from many challenges arisen from the network size and dynamicity, channel resource imbalance, and bandwidth limitations. Other challenges also exist, including instability of peers, the peers’ low participation, large startup and playback delays, low quality of received video, and resource insufficiency in unpopular channels. In this paper, ReDePoly, with some levels of distributed intelligence, is proposed to reduce the bootstrapping delay and to increase the service quality. The key idea is to substitute agents with bootstrapping peers in the channels to model user behaviors and to share the aggregated knowledge among the agents to disseminate the learned models. Accordingly, the agents dynamically predict the behavior of participating peers to pre-assign them to the predicted channels and to parallelize the threads related to viewing behavior and channel switching. In case of congested channels, the new agents are distilled to mitigate the agent responsibilities. Simulation results show that the proposed approach outperforms the other existing methods regarding the channel switching delay, recovery delay, and quality of service.