Energy-Efficient and Delay-Fair Mobile Computation Offloading

Mobile edge computing has been regarded as a new paradigm to achieve mobile computation offloading (MCO), which enables resource-limited mobile devices (MDs) to offload part or all of computation-intensive applications to more powerful computing entities. In this work we study MCO for data partitioned oriented applications. This type of services are usually delay-elastic, but shorter processing delay helps improve the user experience. We consider energy consumption of MDs and latency fairness of the applications. The latency fairness is achieved by allowing multiple offloading MDs to share one channel and allocating a fraction of the channel time for each application, and minimizing total energy consumption of the MDs is achieved by jointly optimizing the offloading ratio, channel assignments, and channel time allocations. For a special case that assigns at most one MD to each channel, a closed-form offloading ratio is derived and optimal channel assignment is obtained through transforming the problem into a weighted bipartite matching. The general problem is then solved by embedding the branch-and-bound iterations with successive convex approximation that solves a series of geometric programming problems. To reduce the complexity, a two-layer recursive method is then proposed that finds the channel assignments, offloading ratio and channel time allocation for each MD. Simulation results demonstrate that the proposed method can achieve low energy consumption and high latency fairness; and much lower energy consumption can be achieved with a slight decrease in latency fairness.