Research And Improve Based On DHT Algorithm In Structured P2P Networks

In recent years, with the fast development of computer networks and communication technology, P2P has become more and more popular because of its outstanding advantages of scalability and fault tolerance. In the short history of P2P applications, it has become one of the main application types that consume a large part of Internet traffic. In the process of taking advantage of the P2P technology, realizing the resource allocation is the key problem.Structured P2P network is the third generation P2P network and its DHT-based search algorithm has the higher search efficiency which is suitable to large-scale P2P networks and its expansibility is very good. It is the study hotspot at present.Firstly, a distributed search algorithms named Kademlia was analyzed and researched deep. Aimed at topology unmatching problem, an improved strategy which separated the nodes into different clusters according to their actual physics information has proposed. Considering heterogeneity of nodes in the network, this paper introduces the cluster header, and chooses high capability node which is cluster header to do more work.Secondly, an improved topology and heterogeneity-based structured P2P touting algorithm named THK is proposed. THK separated the node in the network into different clusters according to their actual physics address vicinity by landmark binning algorithm and RTT technology. Every Cluster chooses a cluster header node and two duplicate of the header node according to the difference nodes capability. Cluster header nodes save the information of nodes in the cluster, in addition, it cache the successful searched information in the cluster and hot resources. Node first searches information in the cluster and cache information. The using of Cache Mechanism has increase the successful probability searching in the cluster. Finally, the performance of THK and Kademlia are compared through experiments. Experiments results show that THK have better performance both in average logical hops and routing latency compared with Kademlia.