节点文献

对等网络中的内容搜索、定位和下载技术研究

The Research on Content Searching, Locating, and Downloading Technologies in Peer-to-Peer Networks

【作者】 陈海涛

【导师】 龚正虎;

【作者基本信息】 国防科学技术大学 , 计算机科学与技术, 2005, 博士

【摘要】 随着Internet的广泛使用以及接入Internet的计算机系统性能的持续增长,一种新的计算模式——P2P计算得到蓬勃发展。P2P计算致力于高效合理地组织和利用Internet边缘上大量分布的计算、存储、通信、信息等资源。从1999年Napster的出现到现在,短短的几年中,P2P应用已经成为成长速度最快,最流行的互联网应用之一,拥有巨大数量的用户。 内容服务是P2P系统的基本功能模块,内容服务应用是P2P技术当前最主要的应用领域。如何在用户广泛分布、数量巨大、节点行为不可控、计算能力和网络连接不均匀的复杂环境下实现高效的内容服务具有重大的研究意义,同时又是具有挑战性的研究课题。 本文的研究聚焦于P2P网络中的内容服务,分析了P2P网络中内容服务的特殊性以及内容服务技术的研究现状,围绕如何在P2P网络中提供灵活、高效、可扩展的内容服务,深入研究了P2P网络中提供内容服务的关键技术,包括内容服务模型的构造、内容搜索技术、内容定位技术、内容下载技术。主要研究工作包括: 1)提出一种基于内容关联的内容服务模型,充分考虑了原始内容和元数据的管理需求差异,在内容空间完成原始内容的管理,侧重解决原始内容的存储、定位和下载;在元数据空间完成元数据的管理,侧重解决元数据的存储、内容关联管理和搜索。利用元数据建立内容之间的关联关系,并给出基于内容关联的内容搜索方法。 2)提出一种基于兴趣相似度的搜索方法——SBIS:在元数据空间通过内容关联关系的发现度量用户之间的兴趣相似度,在具有相似访问兴趣的用户之间建立逻辑连接关系,限制搜索请求优先在具有相似兴趣的朋友节点之间传播。在内容空间为结构P2P网络的情况下,提出SBISS算法实现基于结构P2P网络的兴趣相似度的计算和内容搜索算法。并在SBISS算法基础上提出集成无描述的内容分组方法的SBISSG算法,进一步提高了搜索算法的可扩展性和搜索质量。在内容空间为非结构P2P网络的情况下,提出SBISU算法实现基于非结构P2P网络的兴趣相似度的学习算法和内容搜索算法。SBISU算法根据历史搜索结果被动学习节点之间的兴趣相似度,学习结果稳定而且收敛速度快。仿真实验结果表明SBIS方法稳定高效,在搜索效率和搜索代价上相对于传统算法有数量级的提高。 3)提出一种层次DHT网络的构造方法——HDHT:通过层次式的DHT进一步提高了传统DHT算法的可缩放性;通过对物理邻居节点的聚合降低了网络访问延迟;通过层次式结构的管理优势实现对网络的管理;通过层次管理策略和节点的能力分类提供了一定的服务质量保证;提出一种低冗余的消息广播算法和一种利用并发下载完成大尺寸内容广播的方法。理论分析和模拟实验验证了算法的性能。 4)提出一种基于分组的并发下载方法——GBD:通过内容分组压缩内容分布信息,降低其传播开销、扩大其传播范围;通过节点分组和下载阶段划分的结合优化连接管理,提高邻居节点的内容可用度和连接平均持续时间;采用在考虑自身下载需求的同时考虑邻

【Abstract】 With the wide application of the Internet and the continuously improved performance of the computer systems connected to the Internet, a new computing pattern, Peer-to-Peer computing, is booming. P2P computing devotes to sufficiently and reasonably organize and utilize the Internet edge’s huge decentralized resources of computing, storage, communication and information. Since the introduction of Napster in 1999, P2P applications have become, only in a few years, one of the fastest growing and most popular Internet applications, which hold a huge number of users.Content service is one of the basic function modules of P2P systems. Content service application is one of the most important applications of P2P technology. How to implement efficient content service in a complex environment which contains decentralized and huge-amount users, uncontrollable nodes with unbalanced computing capacity and network connection is a challenging research area and of great importance.The research in this thesis focuses on content service in P2P network. It analyzes the particularity of content service in P2P network and the current research status on content service. In order to provide a flexible, efficient and extensible content service, we did much research on the key technologies for providing content service in P2P networks, including the construction of content service model, content searching techniques, content locationg techniques and content downloading techniques. This thesis makes key contributions to the following areas:1) A content service model based on content correlation is proposed. Based on the different requirements from the management of the original content and metadata, the model manages the storage, locating and downloading of original content in the content space, and it manages the storage of metadata, content correlating and content searching in the metadata space. In this model, the correlation among content is set up by the use of metadata, while the searching of content is based on the correlation of content.2) A search method—SBIS, based on interest similarity is proposed. SBIS computes the interest similarities among users by discovering the content correlations in the metedata space, and constructs logic connections among users with simliar interests. Search requests spread to friend nodes with similar interests with priority. When the content space is a structured P2P network, SBISS algorithm based on the particularity of structured P2P network is proposed to compute the interest similarities and the route search requests. Also SBISSG algorithm based on SBISS is proposed to integrate a new content grouping method without the need of users’ descriptions and improves the extensibility of search algorithm and search quality. While the content space is an unstructured P2P network, SBISU algorithm based on the particularity of an

节点文献中: