A Computation Offloading-based Strategy For Energy Saving On Smartphones

With prosperity of mobile Internet, smartphones provide multi-core processors, sharper screens, larger memory, multiple sensors and radios as well as enormous applications. These together put a heavy burden on battery’s energy consumption. In the meantime, advances in battery technology and energy saving solutions have not kept pace with rapidly growing energy demands. Therefore, energy consumption has always been primary bottleneck for smartphones.In this thesis, we propose a computation offloading-based strategy for energy saving on smartphones. It offloads all or part of computation of an application running on smartphone to the cloud, lets the cloud execute the offloaded computation and then returns execution results to the application. Thus, we can achieve energy conservation and time saving, namely reduce the application’s energy consumption and execution time. Furthermore, it also improves the application’s performance and user’s experience, for example, meet user’s delay-tolerance threshold. The core of our proposed energy saving strategy is a computation offloading decision-making algorithm. Given an application, the algorithm basically compares the following two sets of variables to make the computation offloading decision:the application’s execution time on smartphone and its delay-tolerance threshold, the application’s energy consumption on smartphone and its energy consumption on cloud. In the meantime, we propose the concept of critical transmission energy consumption in the algorithm. Moreover, the algorithm will use following three assisted prediction algorithms:execution time prediction algorithm, CPU workload prediction algorithm and bandwidth prediction algorithm.Finally, we design a system and implement its prototype for the proposed energy saving strategy in this thesis. Furthermore, we conduct two sets of experiments on our system: application experiments and scenario experiments. In the experiments, we analyze the energy consumption and execution time of the applications under different factors. Moreover, experimental results demonstrate the superiority of our proposed energy saving strategy, namely energy and time saving.