%0 Journal Article
%T 实时流媒体P2P可收敛反馈网络结盟算法<br>Alliance algorithm of converging feedback network on P2P real-time media streaming
%A 沈孟如
%A 张晋豫
%J 北京航空航天大学学报
%D 2016
%R 10.13700/j.bh.1001-5965.2015.0300
%X 摘要 客户端贡献的不公平性严重影响系统的服务质量和提供服务的能力,本文提出了一个基于距离汇聚的结盟算法,在动态业务量环境中通过实现可收敛反馈网络,有效消除了随机结盟、内容相似驱动结盟和带宽相似驱动结盟中存在的业务量不合理问题。实现了基于公网IP的静态距离算法和基于探测包的距离动态测量法的混合节点间距离评估机制,较好地解决了探测包测距受业务量波动影响较大、客户端感知测距实时性较差以及IP包测距误差较大的问题。引入了一个具有位置意识的基于Polling的均匀流周期请求协同机制,在保持推—拉周期请求机制开销小优点的同时,提高对抗Serving Peer传输劣化和失败的弹性。仿真结果表明:该机制可以减少业务量不合理和用户不贡献恶意行为的概率,当节点较多时,其能够提供比其他结盟算法更好的时延、丢包率和到达率性能。<br>Abstract：Quality of service and the services providing ability of system are deeply influenced by unfairness of client contributions, a alliance algorithm based on distance-convergence was proposed by implementing the converging feedback network in dynamic traffic condition, and it can avoid the unreasonable traffic which cannot be eliminated by the random-alliance, content-likeness driven alliance and bandwidth-likeness driven alliance. A node mix distance algorithm was implemented, which combines IP static range algorithm and dynamic packet probe ranging method. The node mix distance algorithm can overcome the drawbacks that packet probing range is easily affected by the traffic fluctuation, the real-time feature cannot be ensured by the client perception range, and the IP-based range algorithm lacks precision. Moreover, a uniform flow cycle request collaborative mechanism based on Polling was introduced, which not only keeps the lower-offset merit of the pushing-pulling periodic requesting mechanism, but also provides enough resilience against transport ability decline and transport failure. The simulation results show that our mechanism can reduce the probabilities of unreasonable traffic and the vice behavior of non-contribution, and can provide better performance of delay, packet loss rate and packet arriving rate than other alliance mechanisms in larger scale.
%K 流媒体
%K 对等网络
%K 服务质量
%K 反馈网络
%K 结盟算法&lt
%K br&gt
%K media streaming
%K peer to peer
%K quality of service
%K feedback network
%K alliance algorithm
%U http://bhxb.buaa.edu.cn/CN/abstract/abstract13867.shtml