Stability Prediction And Structural Design Based On P2P Node Behavior

A P2P network is also known as a peer-to-peer network.A node in the network can act as a service node,sending resources or other information to other nodes,or as a normal node,requesting services from surrounding nodes.Because of its scalability and load balancing characteristics,P2P networks are widely used in resource sharing,streaming media transmission,and real-time communication.With the growth of the number of nodes,the traditional fully distributed topology network,centralized topology network and unstructured network all show their own shortcomings,or cannot adapt to the addition and withdrawal of a large number of nodes(churn problem),or robustness poor,or lower searching efficiency.In view of the huge number of nodes and extremely dynamic problems,the P2P network urgently needs to solve the problem of network fluctuations caused by the dynamic nature of nodes,and has developed two directions: Firstly,pick a relatively stable node to act as a supernode by counting the session length of each node;Second,establish a cluster network and locally bind multiple nodes in function to reduce the influence of a single node on the entire network.The main work of this thesis is as follows.(1)By analyzing the social behavior of human society,an aggregation algorithm is designed,which assigns specific attributes to nodes.After new nodes join the network,they will use limited information and attributes to calculate the amount of value obtained,and determine whether to establish a corresponding value based on the amount of value.connect.At the same time,the thesis analyzes that the network generated by the algorithm has a cluster phenomenon,and the cluster has a splitting behavior with the increase of the network size.The algorithm enables network nodes to spontaneously establish local dense connections.Based on this,this thesis designs a resource search mechanism based on the cluster,and at the same time explicitly marks the cluster,and can adapt to the cluster splitting behavior.Simulation experiments show that the mechanism has better performance in dynamic networks.(2)By analyzing the historical data of the node,predict the dynamic change of the node.In this thesis,it is assumed that the session duration of each node has its own relatively fixed range,and a judgment basis is obtained by calculating the expected value of the session duration of the node,and the behavioral mutation of the node may be caused by environmental changes.This thesis proposes the Wb DS algorithm,which uses parameter fitting to make dynamic statistics on the session duration generated by nodes.Experiments show that the algorithm can not only predict the average value,but also has good adaptability to data changes.(3)A structured network is designed to facilitate the position replacement of nodes that may be withdrawn to avoid network jitter.We build a double-chain structure(DCT),which contains a main chain and a sub-chain.The nodes at the front end of the main chain act as super nodes of the chain,saving the location information of each internal node.The threshold is set according to the expected session duration of each node obtained by the Wb DS algorithm,and the threshold is used to determine whether the node performs location replacement.In the simulation experiment,the ratio(ρ)of the number of active exits of nodes in the upper half of the main chain to the number of exits of all nodes is used as the indicator.Since the nodes in the front-end part of the main chain act as super nodes,the smaller the value of ρ,the better the network.Stable.In the chain structure and combined with the threshold setting method,compared with the existing methods,the scheduling mechanism in this thesis has a lower ρ value in different scenarios.