Peer-to-Peer (P2P) systems have attracted much attention in academic community and industry circles due to their promising applications in various domains. This paper presents the authors' research efforts on intr...Peer-to-Peer (P2P) systems have attracted much attention in academic community and industry circles due to their promising applications in various domains. This paper presents the authors' research efforts on introducing complex query capabilities in a P2P environment consisting of numerous peers with large volume of data. An underlying hybrid P2P computing platform, named BestPeer is described first. The connection among peers within BestPeer is selfconfigurable through maintaining the nearest neighbor of peers, and the agent techniques employed in the system ensure its capability of providing sophisticated services. The designs of three P2P data management systems which are all based on BestPeer are described in detail. They provide support for information retrieval, query processing and Web services respectively. Advantages and limitations are discussed, while ongoing work is presented. Current systems can provide basic functions for keyword-based search, SQL-like query processing, and Web services querying and discovery. Some further topics on providing fully-fledged data management functionalities for P2P distributed computing systems with security guarantee are also discussed.展开更多
文摘在分布式Peer-to-Peer (P2P)系统中,每个节点都具有相同的能力,负有相同的责任.它们将自己的资源提供给系统,同时可以共享系统中的信息和服务.但是,大多数P2P系统都具有局限性,比如说Napster,Gnutella,它们提供资源共享的粒度很粗糙,进一步地,它们忽略了文件在内容粒度上的共享.并且,大多数系统中节点的Peer是静态指定的,系统中不支持只具有临时IP的节点加入.提出了一个BestPeer的P2P系统,它具有以下几个特性:首先,它将移动Agent技术与P2P技术相结合,从而可以在数据提供者本地进行数据处理.其次,这个系统具有自配置的特性.一个节点可以动态地决定与哪个节点直接相连(决定哪个节点成为该节点的Peer),从而使网络结构的配置达到最优.另外,BestPeer系统提出了一个位置独立的全局名查找服务器(LIGLO).这个服务器可以惟一地识别具有动态IP地址的节点,从而使节点能够在它的Peer的IP地址发生改变后仍然能够找到它.BestPeer系统建立后,提出了一套评估系统的方法.采用32台Pentium II PC机进行实验,每台机器上都运行基于Java的存储管理程序.实验结果表明,BestPeer系统具有比传统的无自调整能力的P2P系统更好的性能.更进一步的实验还显示了BestPeer系统优于Gnutella 协议下的系统.
文摘Peer-to-Peer (P2P) systems have attracted much attention in academic community and industry circles due to their promising applications in various domains. This paper presents the authors' research efforts on introducing complex query capabilities in a P2P environment consisting of numerous peers with large volume of data. An underlying hybrid P2P computing platform, named BestPeer is described first. The connection among peers within BestPeer is selfconfigurable through maintaining the nearest neighbor of peers, and the agent techniques employed in the system ensure its capability of providing sophisticated services. The designs of three P2P data management systems which are all based on BestPeer are described in detail. They provide support for information retrieval, query processing and Web services respectively. Advantages and limitations are discussed, while ongoing work is presented. Current systems can provide basic functions for keyword-based search, SQL-like query processing, and Web services querying and discovery. Some further topics on providing fully-fledged data management functionalities for P2P distributed computing systems with security guarantee are also discussed.
