Least Popularly Used: A Cache Replacement Policy for Information-Centric Networking

Abstract

The local caching policy of nodes is always one of the hottest issues in caching networks. It can be further divided into two branches, the decision policy and the replacement policy. In this paper, a replacement policy named Least Popularly Used (LPU) is proposed for cache nodes in Information-Centric Networking (ICN) to improve the network performance. Both the request number and the arrival time of a content are taken into account in LPU for judging content popularity. Thus, recently popular contents are cached near consumers, reducing the consumers’ distance to access contents and the traffic load on the original sever. Based on the string, tree and ISPlike topology, extensive simulations are performed, and the results indicate that the performance of LPU is better than that of other replacement policies, including Most Recently Used (MRU), Least Recently Used (LRU), Random, Greedy Dual-Size (GDS), Most Frequently Used (MFU), Least Frequently Used (LFU) and LFU-Dynamic Aging (LFU-DA), in terms of the average hop counts, the server hit ratio and the cache eviction counts.