A Reinforcement-Based Push-Pull Approach for Peer-to-Peer Live Streaming

Existing mesh-push approaches for peer-to-peer live streaming systems have been reported to significantly reduce diffusion latency, overlay- construction difficulty, maintenance complication, and churning impact. However, those approaches incur excessive duplicate data and result in bandwidth inefficiency due to stale information as well as chunk synchronicity problems. In this paper, we propose a reinforcement-based push-pull approach for peer-to-peer live streaming over mesh overlays to overcome the mentioned problems. We design a probabilistic push scheme with a reinforcement mechanism as a part of our hybrid approach. The mechanism encourages non- duplicate-sending nodes to push chunks and discourages duplicate-sending nodes from pushing chunks. As a result, duplicate data is significantly reduced and bandwidth efficiency is tremendously improved. We implement and evaluate our approach on NS-2. The simulation results indicate that our approach can reduce 73% of duplicate data under investigated scenarios.