The Analysis And Improvement Of Resources Locating Model In The Kademlia Protocol

Because of the rapid development of Internet, the number of information and shared documents in the network is on the exponential growth. How to quickly and accurately access valuable resources in the Internet has become a topic that has a growing concern from people.As a key foundation technology of P2P applications, the Resource Location Model of P2P network has been a focus that people research. The Resource Location Model determines the resource consumption, the efficiency and accuracy of searching in P2P network. And it affects the services'quality and availability of P2P systems directly. So the research in Resource Locator Model is of great significance. In many P2P technologies, Kademlia is the most widely used one with a simple and practical principle. And it represents the latest technology in P2P network. So the research in Resource Location Model of Kademlia protocol has more practical significance.This article mainly studies the core issue of P2P technology, such as network topology, routing algorithm, system maintenance, etc. Then this article analyzes and evaluates various P2P protocols from the services'efficiency, the services'accuracy, scalability, reliability, load balancing, resource consumption and other aspects. Through such analysis and research, the main defects of Kademlia protocol are found. They are the neglect of the heterogeneity between nodes, the slow update of k-bucket and the mismatch between physical topology and logical topology, etc.For the defects of Kademlia protocol, this article proposes two improvement strategies on the basis of previous research. The first one is using a "comprehensive performance" model to calculate and evaluate the performance of each node and making the nodes with high performance take on more search request, so as to balance the heterogeneity of nodes. The second strategy is using double k-buckets. One k-bucket is working k-bucket the nodes in which are sorted by comprehensive performance. Then the nodes with high performance are choosed easily to search resource. The other one is reserve k-bucket the nodes in which are sorted by the order of they into the k-bucket. In order to investigate the time-line of new nodes, only the first one in the reserve k-bucket can move to the working k-bucket and receive search request. Using double k-buckets not only considers the performance requirement of search, but also considers the stability requirement of the network. Then the topology mismatch problem of Kademlia protocol is only the selection of network performance. The topology mismatch problem may choose the nodes with low network performance to search resource and neglect the neighbor nodes with small physical hop. Because the network performance has been considered in the "comprehensive performance" model, meaning the neighbor nodes are choosed first when searching resource, the topology mismatch problem is incorporated into the heterogeneous problem and then hey are resolved together.