Abstract: Large-scale storage system based on P2P architecture is proposed to provide reliable and effective storage service. In these systems, the replica of data object is used to improve performance, availability and reliability. When data object changes in one of the replicas, data must be transferred to other replicas. This paper proposes a application layer multicast protocol——PBMB. Node in the system is grouped by BID. Messages are transferred by PBMB using a novel method. PBMB is self-organized, and can be used in WAN environment. Simulation shows that the algorithm is stable even in large-scale system. Messages are transferred effectively while using less node resources. PBMB is suitable for real systems in the Internet.

Key words: P2P network, data replica, application layer multicast, group communication

摘要： 在基于对等网技术构建的广域网存储系统中，保存数据的多份副本可有效地提高整个存储系统的性能与可靠性，为了使更新数据在副本间有效地进行传输，该文提出适用于广域网存储系统副本间数据传输的PBMB算法，通过广播位将系统中的节点划分为不同组，并使用一种新颖的方式在组间进行数据传输，与传统方法相比，降低了节点负担，提高了数据传输效率。模拟实验表明，PBMB算法降低了节点连接出入度，具有较好的数据传输效率，适用于实际系统的需求。

关键词: 对等网, 数据副本, 应用层多播, 组通信

CLC Number: 

• TP301.6