Energy-efficient Autonomic Offloading in Mobile Edge Computing

The booming growth and popularity of mobile devices have led to the surge of various mobile applications. Many mobile applications, such as online video, gaming, are essentially computation-intensive, and hence can quickly deplete mobile devices' battery energy. To address this issue, academia and industry have proposed mobile edge computing (MEC) that can enable mobile devices to automatically offload computations to the edge servers located within the radio access networks of cellular operators. However, energy-hungry wireless communications incur extra energy consumption that may offset the energy saving due to computation offloading. To this end, we design an energy-efficient autonomic offloading scheme by jointly considering the physical layer design and application running latency. Specifically, we first mathematically model the energy consumption of a mobile application in MEC environment by taking into account the energy consumption incurred by the interactions among the tasks for the same application, which is largely ignored by previous studies. Then, we identify task execution flows based on a task interaction matrix, and formulate the maximum of the task flow's latencies as the application's latency. Finally, we formulate an energy-efficient offloading problem, which is generally NP-hard, and develop an efficient heuristic method to solve the problem. We present extensive simulation results to show that our proposed scheme can achieve significant reduction (up to 20% around) in energy consumption compared with previous schemes.