Online intelligent resource management for power-delay tradeoff in backhaul-limited mobile edge computing systems

Mobile edge computing (MEC) has emerged as a promising paradigm to enhance the quality of experience for mobile device users, as it can offload computation intensive tasks from mobile devices to edge clouds. Nevertheless, task offloading also incurs additional overhead in both power consumption and execution delay. In essence, energy consumption and task delay cannot be optimized simultaneously as these two objectives contradict each other. Although many studies have been conducted to make tradeoff between them, the backhaul limitation of MEC-enabled wireless networks has not been tackled. In this paper, we study the power-delay tradeoff for MEC in backhaul-limited wireless networks. An average weighted sum power consumption minimization problem with task buffer stability constraints is formulated to investigate the problem, and a Lyapunouv Optimization based online intelligent computation and communication resource management algorithm is also proposed to solve it. In particular, at each time slot, optimal CPU-cycle frequencies, transmit power and backhaul capacity allocation are efficiently determined. Performance analysis indicates that under the proposed algorithm an [O(1/V), O(V)] tradeoff exists between the power consumption and execution delay, where V is a control parameter. Finally, simulations are conducted to validate the analytic results and demonstrate the changes of the system performance with various algorithm parameters.