Coercion builds cooperation in dynamic and heterogeneous P2P live streaming networks

Abstract Sustaining reasonable performance in a peer-to-peer (P2P) network is contingent upon cooperation among peers. As autonomous agents, peers cooperate only when they are incentivized to do so. Typically, incentives are provided through bilateral exchange of useful data, in a tit-for-tat manner. Unfortunately, good for file sharing and video-on-demand systems as they are, such incentive schemes are ineffective in live streaming systems due mainly to the lack of sufficient opportunity to allow such reciprocity to happen, as recently demonstrated in a pioneering quantitative study. The key insight is that with stringent time constraint, good system performance can only be sustained by judicious peer selection. Despite that some pioneering efforts are done in peer selection, it is as yet a difficult challenge to tackle the inherent non-cooperation of peers in a dynamic and bandwidth-diverse network. In this paper, we meet this challenge by first presenting a novel hierarchical game model, covering strategic interactions among peers, trackers, and the content provider. Based on the analytical insights derived from the repeated game model, we propose a Striker strategy to coerce peers to cooperate, leading to significantly enhanced system performance, as demonstrated by our analytical and simulation results. Most importantly, our proposed incentive schemes are highly practicable in a real-life P2P network.