Research Of Fairness Mechanism In P2P Streaming System

With the rapid development of broadband network technology and multimediatechnology, media streaming systems arebecomingmore and more popular in theInternet. Since the peers in peer-to-peer (P2P) network can not only downloaddatafrom networkbut can alsouploaddata to the network, by introducing P2Ptechnology,the resources of the peers in media streaming system can be effectivelyutilized,thusthe server load can be significantly reduced, which makes the systemmore scalable and practical. Many leading video service providers adopt P2Pparadigm to construct their distribution infrastructure.Due to the stringent need of video chunk availability before deadline, most of therealdeployed large-scale P2P streaming systems exploit upload bandwidth of peers asmuch as possible to improvesystemperformance, whereas the fairness among peershas not been carefully addressed.According to the recentmeasurementresults ofalarge-scale P2P streaming system, namely,PPLive, although enjoying the sameplayback quality, the upload rate of high-bandwidth peers is almost ten times of thatof low-bandwidthpeers, it's extremely unfair for the high-bandwidth peers. Suchunfairness will hinder the incentive mechanism of the system and introduce theproblem of free-riding in the system.We develop an amenable model to study the performance and fairness of P2Pstreaming systems.With the mathematical model, we derive the performance andfairness level of system in the case of achievingoptimal performance, optimal fairnessand optimal fairness under a given streaming rate. We point out the rate assignmentstrategies with which system achieves these different forms of optimality andderivethe tradeoff between performance and fairness. To achieve the optimal fairness under a given streaming rate, aMicropayment-based Progressive Filling(MPF)fairness mechanism is proposed. By introducing micropayment model, we canachieve the desired upload rate assignment of peers effectively.What' more, with thecarefully designed neighbor management strategy and chunk serving strategy, themechanism also provides incentive for cooperation ofpeers. Our work is evaluated byperforming extensivesimulations on the platform of p2pstrmsim, simulationresultsprove that our mathematical model can provide insights into the tradeoffbetween fairness and performance in P2P streaming systems. And withour MPFfairness mechanism, the system can achieve better fairness;at the same time, themechanism also provides incentive to peers in the system.