Research On Mapping System In Name-address Separation Network

With the development of network technology,rich network applications,and massive mobile devices accessing the network,the Internet is facing scalability problems and mobility problems.As one of the trends for future network evolution,the name-address separation network supports the mobility of the network and solves the problem of routing scalability by splitting the dual identity of the IP address into the name identifying the identity and the address identifying the network location.The distributed mapping system is the foundation and core of a name-address separation network.It is responsible for managing the binding,storage,and query of names and addresses,which directly determines the performance of network applications.This thesis focuses on the design structure and query performance of the mapping system in a name-address separation network.Firstly,in order to solve the mismatch between the physical network and the logical network in DHT,a mapping system based on the location-associated Chord is designed.The structured DHT protocol Chord is used to manage the mapping server nodes,physical nodes are added to the routing table of the nodes in the logical network.Network topology information simplify the query process and reduce the query delay.In addition,in order to reduce the update cost in high-mobility scenarios,the method of managing name-address mapping entries at the two levels inside and outside the domain is used to reduce the update range as much as possible to further improve the mapping analysis performance of the system.Theoretical analysis and simulation results verify that compared to LISP-DHT,the average query delay of a location-based Chord mapping system is lower.Secondly,a dynamic inter-domain replica creation strategy based on Q-learning algorithm and an update propagation scheme based on tree structure are proposed.The idea of reinforcement learning is adopted to dynamically adjust the number and position of replicas through the feedback of the network environment,which saves the bandwidth from regularly exploring the network environment and also improves the query efficiency of the system.The update propagation scheme based on the tree structure constructs a tree structure by filtering nodes in a domain that are adjacent to the replica management node.Compared with the Paxos consensus algorithm,it has lower complexity,higher efficiency,and is more suitable for name-address separation mapping systems.The simulation results show that compared with Auspice,the dynamic copy creation strategy based on Q-learning reduces the average query delay by 7.45% when the system load is high,and the update propagation delay based on the tree-based update scheme reduces the update propagation delay by 45.42%.