Efficient Data Access in Cloudlet-based Mobile Cloud Computing

Abstract

The growth in mobile devices and applications has leveraged the emergence of mobile cloud computing, which allows the access to services at any place and extends mobile computing. Usually, the current mobile network consists of a restricting factor in supporting such access because, from a global perspective, cloud servers are distant from most mobile users, which introduces signi cant latency and results in considerably delays on applications in mobile devices. On the other hand, Cloudlet are usually on the edge of Mobile Networks and can serve content to mobile users with high availability and high performance. This thesis reviews both the traditional mobile cloud computing and the Cloudlet architecture. A taxonomy on the Cloudlet architecture is introduced and three related technologies are discussed. Based on the user needs in this environment, personal model which is used to predict individual behaviour and group model which considers caching popular data for several users are proposed. Making use of these two models and the Cloudlet architecture, two data access schemes are designed based on model distribution and data pre-distribution. We have conducted experiments and analysis for both the models and data access schemes. For the models, model efficiency and comparisons among different technologies are analysed. Simulation results for the data access schemes show that the proposed schemes outperform the existing method from both battery consumption and performance aspects.