Research On Media Popularity Prediction And Caching Under Radio Access Network

With the fast development of wireless communication technologies, the 4G mobile networks have come into our lives. It has become quite common for people to use mobile devices to surf the Internet via the mobile networks. This trend has led to an enormous growth of the data traffics of the mobile networks. In order to cope with the problems caused by the big mobile data burdens, such as network congestion and communication failure, caching in the mobile radio access networks is considered as an economic and efficient solution.In modern networks, a large amount of network contents are created and shared every day by ordinary users. There are many issues related to such user-generated contents, including their great amount, their short lifespan and their time-varying popularity dynamics. Traditional popularity prediction methods are incapable to effectively resolve these issues.In this paper, we first focus on the popularity prediction of netwrok contents with time-varying populairty dynamics. Based on the relationship between users’interest variation and the network content popularity variation, we propose an adaptive popularity prediction approach based on a dual process method for the user-requested data. In the first process, we implement the EWMA populrity prediction method to calculate network media files’ popularities. In the second process, we focus on the evaluation of the varying speed of users’interests. Then we use the estimated speed of the variation of the users’ interestes to adjust the key parameters in the EWMA appraoch, which can keep the cache system’s popularity tracking speed to be in pace with the network hot contents’ shifting speed and cache the "current" hot contents.Moreover, in order to reduce the number of redundant replicas of the same content on cache nodes which are adjacent to each other and make better use of their cache space, we propose a cache content sharing approach for modile clients in the mobile base station environment. In order to further improve the probability of finding concerned contents on neighbour mobile clients, we cluster and associate users with similar content preference to the same base station. Therefore we can efficiently reduce the chance to share cache space amoung users with no common insterests and yield cache sharing among users with similar content preference. For the purpose of reducing redundant replicas of the same content, we propse a centralized cache management approach. This appraoch takes the cache space of the mobile clients as a whole cache pool. When cache space is needed, it first deletes some replicas of the file which own the most replicas in the cache pool to free space. When the number of replicas for every cached file is one, it uses popularity comparision to determine whether to cache the new content or not. In this way, it makes a better use of the mobile clients’cache space.