中文摘要：

传统的基于P2P的实时流传输系统中的成员管理机制有的依赖于全局成员信息，缺乏可伸缩性；有的采用随机的节点选择策略，很难获得优化的全局性能．本文提出SEBMM（Synchronization Estimation based Membership Management，基于同步评估的成员管理）机制，将对等节点的选择与网络性能和节点之间的数据同步性相关联，同时建立了一套全分布式动态节点推荐机制，使系统对网络的动态改变具有很大的弹性，并潜在地帮助系统收敛于合理的全局性能．系统模型的仿真显示了系统的正确性和有效性．

英文摘要：

Traditionally, some membership management schemes rely on global membership information, which is lacking in scalability; the others adopt a random peer selection scheme, which can not be easily modified to gain perfect global performance. This paper presents SEBMM （Synchronization Estimation based Membership Management）, which correlates the peer selection to network performance and data synchronization between peers. A distributed dynamic peer suggestion scheme is also proposed, which is resilient to network dynamic and can potentially help the system to converge to a reasonable performance. The validity and effectiveness of this solution are demonstrated in simulations.