Design And Implementation Of High Performance Mapping System Of The Identity/Locator Separation Network

The Identity/Locator Separation Network split the identity and the location.In Identity/Locator Separation Network,the identity space and location space represent the identity and location of the nodes.The mapping system,which serves as a significant part of the Identity/Locator Separation Network,takes charge of the storing and managing of the identity map and provides the mapping query service.The structure and performance of the mapping system have a huge influence on the mapping resolving delay,which,as a result,affects the performance of the Identity/Locator Separation Network.This work focuses on reducing the resolving delay of mapping from two aspects.On the one hand,we propose the mapping prediction algorithm for fixed identity to improve the hit rate of the local cache and to reduce the delay of mapping resolving;on the other hand,by tracing the movements of mobile node,we put forward the location prediction algorithm to reduce the delay caused by switching.The main innovations are as follows:1.We proposed a mapping prediction algorithm for fixed identity based on collaborative filtering(CFCIP).The frequency of the access to fixed identity is collected by the edge router,and then mapping system make prediction based on collaborative filtering for edge router,and push the predictions to edge route actively.The experiments show that,combined with the replacement strategy LRU or LFU,the convergence time from the initial state to the stable state of the proposed algorithm is three times faster than that of the replacement strategy LRU or LFU.When cache is in a stable state,the hit rate of the proposed algorithm achieves 80%.It is 8% higher than the replacement strategy.2.We propose the mobile identity prediction algorithm based on mobile behavior similarity clustering(UCMBSP).Based on mobile behavior similarity clustering,the mapping system make position prediction,form the mobility identity mapping table,and then push the table to the edge route.The experiments show that when predicting one position,the accuracy of the algorithm is 68%.It is 13% higher than EdgeBuffer.3.Based on the prediction algorithm above,we design and implement the IDMapServer,which is an Identity/Locator Split network mapping prototype system.Our test results proved the validity of the algorithms.