Towards energy efficient task execution in mobile cloud computing

Mobile cloud computing has been touted as an effective solution to extend capabilities of resource-poor mobile devices by application offloading. With the emergence of cloud computing, various cloud-assisted mobile application platforms have been proposed, such as Cloudlet, CloneCloud and Weblet, which leverage cloud resources to accomplish computation intensive tasks and save energy consumption on mobile devices. Nevertheless, it is not always energy efficient to offload mobile applications to the cloud for execution. The stochastic nature of wireless channel and various profiles of mobile applications (e.g., task topology and time deadline requirement) present challenges for making decision on application offloading. In this thesis, we investigate energy efficient task execution in mobile cloud computing. First, we aim to minimize the energy consumption on the mobile device while meeting the time constraints for the execution of all the tasks under stochastic wireless channel. Specifically, we consider three cases in terms of the task topology within mobile applications, i.e., one node, linear chain and general topology. For the first case, the whole application is executed either on the mobile device (i.e., mobile execution) or on the cloud (i.e., cloud execution). We obtain a threshold and find an operational region to determine which execution is more energy efficient. Simulation results show that cloud execution consumes 4.65 times energy less than mobile execution for some cases. For the second case, each task in the linear chain is sequentially executed, with output data as the input of its subsequent task. We propose a collaborative task execution between the mobile device and the cloud. We formulate the collaborative task execution as a constrained shortest path problem and derive a one-climb policy, which indicates that the energy-optimal execution only migrates once from the mobile device to the cloud if ever. Simulation results show that more than 5 times of energy consumption can be saved by the collaborative execution. For the third case, a mobile application consists of fine-grained tasks in general topology. We formulate the collaborative task execution as a delay-constrained workflow scheduling problem. We leverage Partial Critical Path analysis and adopt one-climb policy to schedule the task execution. Simulation results show that the proposed collaborative execution can save energy consumption compared to the local execution on the mobile device and is more flexible than the remote execution to the cloud. Then, we aim to minimize the energy consumption in the cloud for transcoding as a service (TaaS) and virus scanning as a service (VSaaS). For the former, tasks of video transcoding can be executed locally on mobile devices, or offloaded to a set of service engines in the cloud. The objective is to reduce the energy consumption on both mobile devices and the cloud for executing transcoding tasks. For the mobile device, we find an operational region to determine whether the task should be offloaded or not. For the cloud, we leverage Lyapunov optimization framework and propose an online algorithm to reduce energy consumption on service engines while achieving the queue stability. By choosing the control variable, one can achieve a time average energy consumption arbitrarily close to the optimal solution. Simulation results show that the proposed algorithm can outperform two alternative algorithms. For the latter, tasks of virus scanning on mobile devices are offloaded to the cloud, which supports N-version protection to allow one file to be scanned by multiple distinct service engines in parallel for less missed detection error cost. Similarly, under Lyapunov optimization framework, we propose an online algorithm for task scheduling to reduce the energy consumption of service engines and the missed detection error cost, while achieving the queue stability. We show that the proposed algorithm is energy-efficient to provide N-version protection for VSaaS. The proposed task execution policy, including collaborative task execution and task scheduling algorithms, can reduce the energy consumption on mobile devices and the cloud, which provides guidelines for the design of green mobile cloud.Doctor of Philosophy (SCE