非结构化对等网络资源搜索方法研究-智能与分布计算实验室

非结构化对等网络资源搜索方法研究

姓名	彭俊杰
论文答辩日期	2007.01.26
论文提交日期	2007.02.05
论文级别	硕士
中文题名	非结构化对等网络资源搜索方法研究
英文题名	A Study On Search Method In Unstructured Peer-to-peer Network
导师1	卢正鼎
导师2
中文关键词	非结构化对等网络;搜索;核心节点;边缘节点;朋友节点
英文关键词	Unstructured Peer-to-Peer Network;Search;kernel-node;edge-node;friend-node
中文文摘	作为一种全新网络拓扑结构，Peer-to-Peer（P2P）网络打破了传统的客户机/服务器模式，网络中每个节点地位都是相同的，在为其他节点提供服务的同时也享用其他节点提供的服务。P2P为实现网络资源的全面共享和合作提供了可能，并在文件共享、对等计算、搜索引擎、协同工作等众多领域中显示出巨大的优势，尤其是在文件共享方面的应用十分广泛。 P2P网络主要可分为集中目录服务器模式，分布式非结构化网络模式以及分布式结构化网络模式。其中集中式网络中用于辅助资源定位而使用的目录服务器容易成为系统的瓶颈；结构化网络难以有效地支持模糊搜索；非结构化P2P网络因其简单和健壮性而获得广泛应用，目前使用最多的文件共享软件都基于非结构化网络的“扩散”的搜索的方法。随着网络规模的增长，不仅搜索消息的数量在增长，同时由每一条消息产生的潜在流量也在大幅增长，浪费了大量的网络带宽和计算能力。因此，应该研究和改进非结构化P2P网络的资源定位机制。针对上述问题，对非结构化P2P网络中资源的搜索策略进行改进，包括考虑P2P网络中节点的异构性，综合分析节点的查询能力、存储能力、传输资源的能力以及处理信息的能力，给出节点划分的准则，将节点划分为核心节点和边缘节点，让核心节点在资源搜索的过程中发挥更大的作用，并从节点以往查询历史中总结发现节点的兴趣，进而为每个节点引入若干朋友节点。将资源搜索过程分为三个阶段：通过访问朋友节点快速搜索，通过搜索核心节点访问到系统中的绝大部分资源，通过访问边缘节点搜索保证算法能够搜到系统中共享的每个文件。若在上述搜索过程任意一个阶段查询获得足够的返回结果，则查询过程结束，否则转入下一阶段。最后通过模拟测试表明改进算法稳定有效，降低了网络开销。
英文文摘	A new form of network called Peer-to-Peer network has come into being. P2P network breaks the traditional model of Client/Server. Each node not only plays a role of server that provides the service but also plays a role of client that enjoys the uses of resources in Peer-to-Peer network. Peer-to-Peer network make it possible that spare computer powers can be taken full advantage of. The precondition of sharing resources in large area is to route the resources efficiently. The new pattern of network brings some new characters such as no center in the network, so it more difficult to search the resources efficiently. There are already some methods in searching resource in P2P network. Napster uses a central server that stores the indexes of the files to help finding the resources, but the central server will become the bottleneck. Decentralized but Structured models use Distributed Hash table (DHT) method, can find resource in few hops, but it doesn’t support the fuzzy search. Unstructured P2P network is widely used because it is simple and robust. Most of the file-sharing software in P2P network uses the flooding method to search resource. With the growth of the network, the number of the message increases rapidly. These messages include a lot of unnecessary duplication of packet flow. Therefore, we should improve the search mechanism which is simple and efficient. This paper introduces some improved search approaches, including taking into account the heterogeneous nature of the P2P network node and dividing nodes into two types: kernel-nodes with more resources and edge-nodes by their answer capability, storage capacity, resources and the ability to handle information transmission。Discovering the interest of each node from their query history and introducing friend-nodes to each peer. Dividing the search process into three phases: searching Friend-node, Kernel-node and Edge-node. If at any stage of the query return enough results, the search process will be completed. Finally, the Simulation shows that improved methods enhance the search efficiency and reduce the network costs.