Caching algorithm for content-oriented networks using prediction of popularity of contents

It is important to use cache efficiently for the content deployment in the aspect of reducing load of server and latency, especially in content-oriented networks such as ICN (Information Centric Networking). Since the capacity of cache on each network node is limited, numbers of cache replacement algorithm have been proposed. However, because of previous methods do not consider the rapid fluctuation of content demand, ineffectual cache of content remains which cause a waste usage of cache capacity. In this paper, we propose the method of using the prediction result of content demand. We first demonstrate that by applying prediction can improve the efficiency of cache utilization. However, predicting the demand of whole contents is not realistic in the aspect of computational cost. We therefore propose the method of reducing computational cost greatly without lack of cache efficiency. We demonstrate that by simulating with real monitored data, cache hit rate can be increased about 1.6 times. Furthermore, we show that our method has advantage in many aspects, such as efficient usage of cache and content deployment based on its popularity in the whole network by comparing with other existing methods.