Abstract: Super-peers P2P network dividing peers into super-peers and leaf-peers sufficiently settled the problem of low resource locating efficiency due to the coexistence of heterogeneous peers.However，traditional super-peer P2P network does not adapt effective methods to build an efficient super-peer layer and adjust super-peers layer according to the high dynamic network situations.A novel efficient super-peer P2P network construction algorithm called RSSN is proposed in this paper.RSSN is based on the random walk sampling algorithm，it chooses more capable peers for preliminary super-peers that can dynamic adjust the super-peers layer according to network conditions. Additionally，through caching index of sharing files in preliminary super-peers，RSSN is capable to adapt to frequent dynamic activities.The simulation evaluation shows that RSSN can dramatically improve the efficience and robust performance of the super-peers P2P network compared with Gnutella0.6.

Key words: super-peer, preliminary super-peer, leaf-peer, random walk algorithm

摘要： 超节点对等网的引入，有效解决了网络节点异构性所带来的低性能节点对于文件定位效率低的问题。但是传统超节点对等网构建效率低，不能适应目前高度动态的网络环境。提出一种高效可靠的超节点对等网RSSN，RSSN通过漫步算法对网络叶节点采样，从采样集合中选出高性能节点建立预备超节点，通过判断网络需求调整超节点层，并利用预备超节点备份文件索引信息，提高对等网的稳定性。仿真实验表明，相较Gnutella0.6超节点对等网，RSSN能够有效地提高对等网中超节点的平均性能和利用率，并能适应高动态的网络环境。

关键词: 超节点, 预备超节点, 叶节点, 漫步采样算法