J4 ›› 2012, Vol. 39 ›› Issue (3): 136-143.doi: 10.3969/j.issn.1001-2400.2012.03.022

• 研究论文 • 上一篇    下一篇

P2P流媒体直播系统自适应邻居节点选择算法

冯侦探1,2;倪宏2;王劲林2;尤佳莉2
  

  1. (1. 中国科学院 研究生院,北京  100049;
    2. 中国科学院 声学研究所 国家网络新媒体工程技术研究中心,北京  100190)
  • 收稿日期:2011-03-25 出版日期:2012-06-20 发布日期:2012-07-03
  • 通讯作者: 冯侦探
  • 作者简介:冯侦探(1983-),男,中国科学院声学所博士研究生,E-mail: spyfeng@gmail.com.
  • 基金资助:

    863计划资助项目课题“融合网络业务体系的开发”(2011AA01A102);

Adaptive neighbor selection method for the  P2P media streaming system

FENG Zhentan1,2;NI Hong2;WANG Jinlin2;YOU Jiali2   

  1. (1. Graduate Univ. of Chinese Academy of Sci., Beijing  100049, China;
    2. National Network New Media Eng. Research Center, Inst. of Acoustics, Chinese Academy of Sci., Beijing  100190, China)
  • Received:2011-03-25 Online:2012-06-20 Published:2012-07-03
  • Contact: FENG Zhentan

摘要:

针对P2P网络中节点的异构性,提出了一种基于节点动态服务能力的自适应邻居节点选择算法.首先,为分析节点的动态服务能力特性,对流媒体直播系统进行建模,并根据该模型将系统节点按照服务能力分为不同的级别.再采取随机行走的方式选择邻居节点,随机行走过程结束时的节点被加入备选邻居列表.为保证备选邻居节点拥有较强能力,根据节点服务能力动态调整随机行走的期望静止概率分布,采用Metropolis-Hastings算法计算转移矩阵以满足期望静止概率分布.详细描述了节点加入、退出过程以及邻居节点更新策略,以确保节点负载的均衡及系统稳定性.实验结果表明,该算法能显著提高系统性能,降低系统延时,同时对动态网络环境有良好的鲁棒性.

关键词: 邻居节点选择, 随机行走, 覆盖网, 流媒体直播

Abstract:

Taking into consideration the heterogeneity among peers, an adaptive neighbor selection method is proposed based on peer dynamic capacity. Firstly, in order to analyze the dynamic capacity, a model for the P2P streaming system is presented. According to the model, peers are divided into different levels based on their capacities. Secondly, a random walk process is applied for the purpose of searching the neighbor. The peer is added to the candidate neighbor list when the random walk process ends. In order to guarantee that the candidate neighbors have a high capacity, the expected stationary distribution for the random walk is adjusted according to the peers' dynamic capacity. The transition probability matrix is calculated by the Metropolis-Hastings methods in order to satisfy the expected stationary distribution. Moreover, the neighbors' updated method,  and join and departure process are elaborated to ensure the node load balance and system stability. Simulation results demonstrate that the algorithm can significantly improve system performance, reduce system latency, and is very robust in dynamic network environment.

Key words: neighbor peer selection, random walk, overlay, media streaming

中图分类号: 

  • TP393