Research On Cache Replacement Strategy Based On Content Popularity In ICN

In Information Centric Networking(ICN),the efficient deployment of cache content in the network can improve network performance,reduce response time,and reduce the load of the source server,but the space of the router is limited.Therefore,the design of cache replacement strategy is a key to the research of ICN.The greedy caching algorithm is a solution proposed by Banerjee in 2018.Compared with random caching,the caching performance has been improved.The algorithm takes into account the needs of each node,and represents the local popularity based on the cache request flow from the downstream,determines the cache deployment of each node,and optimizes the content cache in the network.However,this method does not consider the interaction of cached content between nodes,nor does it mention the update of content popularity.Aiming at the above two deficiencies,this thesis proposes a cache replacement strategy based on content popularity.During the content request process,the router calculates the popularity and classifies the content.Then,as the data packet returns,caching executes,and hot content are communicated between nodes.In this thesis,the strategy is implemented in ndnSIM simulator,and compared with greedy cache scheme.The results show that the network performance of this strategy is higher than greedy caching in the stable stage,with an average of about 2 percent delay improvement and 8 percent cache hit ratio improvement.Then this thesis analyzes the content requests in the network from different applications.Comprehensively considers several content requests such as video data representing a long access cycle and news data representing a short access cycle,and finds that in the design of the cache replacement strategy,current cycle information,combined with part of the historical cycle information,is the most appropriate to evaluate the popularity of the content.Through the simulation of mixed data on the Visual Studio platform,the experimental results show that in the multi-cycle popularity calculation,the parameter r of 0.7-0.9 is more appropriate,which can balance the accuracy of the popularity calculation and the sensitivity of the cache replacement,will cache popular content in the node.