| The birth of the P2P computing bring strong impetus to the development of data integration.Rely on the server is one of the shortcomings of the traditional solution.. The P2P data integration system combines the advantages of P2P technologies and data integration,it can overcome this shortcoming and provide users rapidly and precisely service of querying.Thinking of numbers of the P2P network based on DHT,Kademlia is widely used as a protocol which is proven and efficient.It has a very clear logical structure which like the binary tree in the data structure.Every node in the network will be allowed an unique ID based on DHT.Kademlia use XOR metric in computing the distance of two nodes.it can provide θ(log") lookup to locate the node closest to a given key in this way.But Kademlia is not perfect,the problems we want to solve in this paper inclue:the problem of mismatching between the logical topology and physical topology,how to reduce the impact to the network itself which is caused by the hot resources,how to improve the speed and accuracy of querying.The main content in this paper is to research these problems and to resolve them.In order to solve these problems,we do some modification on the architecture of node ID with adding the network number of IP address in it.Through the particular XOR metric for distance in Kademlia we can carry our point of matching logical topology ande physical topology. Thinking of the model that we improved,we put forward the concept of ultra node.The ultra node is on the charge of managing nodes and controling the query routing.Thinking of the hot resource,we add fast table in the structure of node to reduce the impact to the network.Firstly,we make some improvement on the structure of node that we regard the network number of IP address as the high-order bit.In this way we make the nodes which are closed on physical close to each other on logical combine the particular XOR metric for distance in Kademlia.Secondly,thinking of the logical topology of the network that we make improvement,we put forward the concept of ultra node.The ultra node only play role in this LAN that it belongs to and it on charge of managing the structure of LAN and the information of nodes.And also ultra node will control the query routing on this LAN.Through the cooperation of the ultra node and ID that we improved,we makd full use of the resource of computers on the network and improve the speed of locating resource largely.Finally, we add fast table in the structure of node thinking of the phenomenon of the hor resource which is offen occur in the P2P network.The main function of the fast table is to record the information of hot resource.With the help of fast table,on one hand we improve the speed of visiting the hot resource,on the other hand we can reduce the impact to the network which is cased by the visiting to the hot resource in a short time. |
