Churn-proof Wireless Caching with D2D Communication

Popular data objects will be cached close to end-users in future wireless networks. Both distributed storage and inter-user communication in wireless systems alongside caching can be utilized to increase the performance of such networks compared to traditional server-client based approaches. This holds true especially when users are clustered in certain geographical areas forming local caching communities, and when user equipment can be used to store and distribute data, which decreases the backhaul load and increases overall energy-efficiency. Energy-efficiency is a key performance metric in modern wireless networks where enormous amounts of information must be transmitted between a large number of devices. The research of this thesis contributes to the study of wireless caching with inter-user communication especially in terms of energy-efficiency. The aim is to minimize the expected overall energy consumption of the cellular system. The joint use of Device-to-Device (D2D) communication and erasure coding for Distributed Storage Systems (DSS) is shown to protect cached data against mobile churn, which increases the energy-efficiency of the system. This is mainly due to the fact that erasure coding provides protection against data loss in caching communities, which in turn decreases the need of contacting remote base stations. Besides coded wireless D2D caching, the statistics of caching for maximal byte hit rates as well as inter-user communication with message forwarding have been studied in detail. The statistical approach to cache sizing takes both file popularities and file sizes into account when designing the cache so that the expected amount of traffic from the cache is maximized, thus alleviating the strain on the origin server. The multihop message forwarding technique decodes functions of transmitted messages so that the probability of a decoding failure vanishes. Numerical simulations are used to verify the theoretical calculations. Both theoretical calculations and simulation results indicate that the use of erasure coding has the potential to drastically decrease the energy consumption when appropriate coding methods are used.