基于Kademlia的P2P资源定位研究

发布时间：2018-02-05 02:15

本文关键词： P2P数据集成 DHT Kademlia网络快表　出处：《山东大学》2014年硕士论文　论文类型：学位论文

【摘要】：P2P技术的诞生给数据集成领域带来了强大的发展动力。传统的集中式数据集成系统具有系统性能依赖服务器这一系统短板,结合P2P和数据集成系统两者优势而诞生的P2P数据集成系统,能够克服传统集中式数据集成系统的这一缺陷,为用户提供高速精确的数据查询服务。在众多的基于DHT的P2P网络中,Kademlia网络作为一种比较成熟且比较高效的协议,得到了广泛的应用。它的逻辑结构类似数据结构中的二叉树,清晰而明了,每个节点根据DHT分配一个独一无二的节点ID,对于计算节点之间的距离,Kademlia网络采用的独特的XOR计算方式。通过这种方式,对于资源关键字,它定位的时间复杂度是θ(logn)。但是Kademlia网络并不是完美无缺的,在本文中我们着重要解决的问题包括：Kademlia网络的逻辑拓扑和物理拓扑不匹配的问题,如何最大限度的减少热点资源对网络本身带来的冲击,如何在原有的基础之上进一步提高资源定位的速度和精确度。对于这些问题的深入研究和解决,成为了本文的主要内容。为了解决这些问题,在本文中,我们对网络中的节点ID进行了一系列修改,将节点的IP地址的网络号加入到了节点的ID之中,通过Kademlia特有的计算节点距离的方式,可以使得网络在物理拓扑上与逻辑拓扑相匹配。针对我们改进的模型,我们提出在原有的节点基础之上增加超级节点的概念,超级节点负责管理本局域网的节点并且担负路由调控功能。针对热点资源,我们在节点的结构中增加了快表来缓解对网络的冲击。首先,对节点的ID结构进行了改进,将节点IP地址的网络号作为ID的高位部分,结合Kademlia固有的XOR计算方式,使得在物理上靠近的节点在逻辑上也处于相近的位置。其次,针对改进后网络的逻辑拓扑结构,为了能够进一步提高资源查询的速度和精度,我们提出在原有的基础上增加超级节点的概念,超级节点只在本LAN中发挥作用,负责掌握本LAN的结构和其中节点的信息,在路由转发过程中超级节点负责本LAN的路由调控职责。超级节点机制和改进的节点ID互相配合,充分调用了网络中的计算机资源,并且在很大程度上提高了资源定位的速度。最后,针对P2P网络中经常出现的热点资源现象,我们在节点的结构中加入了新设计的快表,用快表来记录热点资源的信息。在快表的帮助之下,一方面提高了节点对热点资源的查询访问速度,另一方面也降低了短时间内对热点资源的海量访问所带来的对网络的冲击,可谓一举两得。
[Abstract]:The emergence of P2P technology has brought a powerful impetus to the data integration field. Traditional centralized data integration system has the system performance dependent server, which is a short board of the system. P2P data integration system, which combines the advantages of P2P and data integration system, can overcome the shortcoming of traditional centralized data integration system and provide users with high-speed and accurate data query service. Among many P2P networks based on DHT, Kademlia network is a mature and efficient protocol. Its logical structure is similar to the binary tree in the data structure. Each node is assigned a unique node IDD according to DHT to calculate the distance between nodes. Kademlia network uses a unique XOR computing method. In this way, for resource keywords. The time complexity of the location is 胃 gng, but the Kademlia network is not perfect. In this paper, the important problems we solve include the mismatch between the logical topology and physical topology of the: Kademlia network, and how to minimize the impact of hot resources on the network itself. How to further improve the speed and accuracy of resource location on the basis of the original is the main content of this paper. In order to solve these problems, in this paper, we made a series of changes to the node ID in the network, adding the network number of the node's IP address to the node ID. The network can match the logical topology in physical topology through the unique way of computing node distance in Kademlia. We put forward the concept of adding super nodes to the original nodes. The super nodes are responsible for managing the nodes of the local area network and the routing control function. We have added fast tables to the node structure to mitigate the impact on the network. First of all, the structure of node ID is improved. The network number of node IP address is taken as the high part of ID, and the inherent XOR calculation method of Kademlia is combined. The nodes that are physically close are logically close to each other. Secondly, in order to further improve the speed and accuracy of resource query, we propose the concept of adding super nodes to the improved logical topology of the network. Super nodes only play a role in this LAN and are responsible for mastering the structure of this LAN and the information about the nodes therein. In the process of routing forwarding, the super node is responsible for the LAN routing regulation and control. The super node mechanism and the improved node ID cooperate with each other, which make full use of the computer resources in the network. And to a large extent improve the speed of resource positioning. Finally, in view of the phenomenon of hot resource in P2P network, we add the newly designed fast table to the structure of the node and record the information of the hot resource with the help of the fast table. On the one hand, it improves the query access speed of nodes to hot resources, on the other hand, it also reduces the impact on the network caused by massive access to hot resources in a short time, so it can be said that killing two birds with one stone.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02

【参考文献】