Research On Cache Management Technology Of Information-Centric Networking

The birth of the Internet has changed people's way of life and production,which has affected human progress in all aspects.However,the traffic in the network is increasing year by year,and the proportion of multimedia data is also increasing,there are a lot of redundant transmission in IP network,reducing the efficiency of network,gradually became the bottleneck of network development.The traditional network developed from the telephone network is mainly a connection-oriented network which is responsible for end-to-end transmission.Its main purpose is to share resources between two static hosts.However,nowadays,this communication mode has gradually developed to share information and data in the network.The proposal of Information-Centric Networking effectively improves a lot of disadvantages of traditional IP network.It decouples the relationship between information and location,shares data in the network and does not care about the location of data.In the Information-Centric Networking,the router has the caching function,which can cache the passing data.When the next request with the same name passes through the router,the router will directly respond,thus reducing the length of network transmission.Therefore,the caching strategy directly affects the performance of the Information-Centric Networking.The research on caching strategy can be divided into two directions.One direction is caching replacement policy.When the caching queue is filled up,it is necessary to decide whether to replace and which cache to replace through caching replacement policy.The other direction is the caching placement strategy,which determines the caching node or nodes to place.In this paper,a dynamic adaptive pre-filter queue cache replacement algorithm is proposed,which can adapt to different network conditions.Based on this,a cache placement strategy based on cache node status and location is proposed.The corresponding evaluation nodes are used to select the appropriate nodes for caching.Firstly,the caching replacement policies of FIFO,LRU,RANDOM and LFU are analyzed in this paper.FIFO caching replacement algorithm ignores the impact of cache content popularity on cache hit rate,LRU caching replacement algorithm has poor cache effect on periodic access content that exceeds the length of cache queue.RANDOM caching replacement algorithm has randomness,and its performance is very unstable when the network situation is complex,LFU cannot perform within the time complexity of O(1).Therefore,there is still room for further improvement of caching replacement algorithm.According to the advantages and disadvantages of each caching replacement policy,this paper proposes a dynamic LRU-K caching replacement policy.By using the pre-filter queue,the cache accessed infrequently is filtered.The number of pre-filter queues is constantly adjusted according to the hit rate of the cache queue and the filling factor of the pre-filter queue,so as to improve its overall hit ratio.The simulation results show that the dynamic LRU-K caching replacement policy can effectively improve the cache hit ratio.Caching placement strategy research is divided into three categories: non-cooperative cache,explicit cooperative cache and implicit cooperative cache.Explicit cooperative caching can generally achieve the best caching effect,but it requires a lot of communication to satisfy the exchange of cached node information and relies on the controller for calculation.Therefore,an implicit cooperative caching placement strategy of on-path is proposed in this paper,based on the dynamic LRU-K caching replacement policy,the nodes with the best state value on the transmission path are selected for caching according to the comprehensive evaluation of the network location and state of the cache nodes.The state value of each cache node is calculated with different weights based on the number of pre-filter queues K,the hit position of the prefilter queue and the distance from the request node on the link.Each request selects one or two nodes in the request path for caching,which reduces the redundancy of data.Simulation results show that the caching placement strategy based on cache node state and location can improve the hit rate and reduce the average request length by a small margin.