An Improved Kademlia-based Resource Locating Model

The network resources storaged in P2P network are increasing greatly.The way how to accurately locate the resources needed in the massive network resources (resource location model of P2P) has become a hot spot in the field of P2P research domain. The accuracy of resource searching and the efficiency of resource transaction are determined by resource location model of P2P. Kademlia resource locating model as the latest technology of P2P network is most widely used and with a simple and practical principle. Therefore, it has great significance to study the Kademlia resource locating model.Firstly, the article mainly introduces the P2P technology, the DHT and the Kademlia. Through the analysis of Kademlia, Some problems are found. Such as the problem of heterogeneity between nodes, the mismatch of physical topology and and logical topology, the problem of the slow update of k-bucket, the lack of reputation mechanism.Secondly, solutions are proposed to solve the problems of Kademlia one by one. For the problem of heterogeneity between nodes, the capability model based on the surplused computing power, the surplused memory and the continuous online time is proposed. For the mismatch problem of physical topology and logical topology, the distance model based on the communication cost between nods got from P4P-distance interface, the RTT and TTL is proposed. Then, dis_cap model based on capability model and distance model is proposed. For the problem of Lack of credibility mechanism, FCircle model is proposed based on the concept of friends, in this model, the credibility value of target node is computed by the Maximum Likelihood Estimate and the resource transaction record. For the problem of the slow update of k-bucket, a new updating strategy is proposed. The structure of k-bucket is readjusted, The credit value is divided into four block, the nodes in k-bucket is ranked from the first block to the fourth block, Then the nodes in every block of credit value is ranked from the maximum dis_cap value to the minimum dis_cap value. This structure of K-bucket can ensure the node with higher credibility value and dis_cap value is closer to the current node, which improves the probability of being selected.Finally, the improved model is tested through OverSim simulation platform, it turns out that the improved Kademlia model greatly improves the success rate of resource locating and the efficiency of resource transmission.