An energy-efficient, time-constrained scheduling scheme in local mobile cloud

Mobile devices have limited resource, such as computation performance and battery life. Mobile cloud computing is gaining popularity as a solution to overcome these resource limitations by sending heavy computation to resourceful servers and receiving the results from these servers. Local mobile clouds comprised of nearby mobile devices are proposed as a better solution to support real-time applications. Since network bandwidth and computational resource is shared among all the mobile devices, a scheduling scheme is needed to ensure that multiple mobile devices can efficiently offload tasks to local mobile clouds, satisfying the tasks' time constraint while keeping low-energy consumption. Two critical challenges need to be solved: (1) estimation of the energy consumption and completion time for tasks to be scheduled, (2) schedule the tasks from multiple source nodes to an appropriate device to accomplish the computation and receive the results.;In this thesis, the adaptive probabilistic task scheduler for local mobile clouds is proposed. The scheduler relies on periodic network messages to discover neighboring computation and network resources. It first estimates the completion time and energy consumption at each potential processing node. Next, it schedules the current task to the proper processing node in a probabilistic way and adaptively adjusts its time margin to improve performance under the unpredictable network condition. Comparing with other existing scheduling schemes, the experimental results confirm that the proposed scheduler achieves highest task completion rate and the lowest average energy per successful task. In addition, the proposed scheduler is able to accommodate different types of tasks and network scenarios.