A Multi-Layer Offloading Framework for Dependency-Aware Tasks in MEC

Abstract

Mobile Edge Computing (MEC) is a promising solution to reduce the task execution delay by placing the computation resource at the network edge close to the end-users, and has received an extensive attention in the 5G era. In this work, we study a multi-layer task offloading framework for MEC, where each task generated by a mobile device can be offloaded to other mobile devices via D2D links, or edge servers with cellular links, or remote cloud server via Internet. We consider a generic task model, where each task can be divided into a set of dependent subtasks and each subtask can be offloaded to different locations. In such multi-layer offloading framework with dependence-aware tasks, we are interested in the optimal subtask offloading problem for mobile devices, that is, how to optimally offload the subtasks of all devices. To study this, we formulate an Energy Consumption Minimization problem for mobile devices, which decides when and where each subtask will be scheduled, aiming at minimizing the total energy consumption of mobile devices. The problem is challenging due to the non-convex constraints. We propose some mathematical operations to relax the nonlinear constraints into linear constraints, and hence transform the original non-convex problem into a linear programming, which can be solved efficiently. Simulation results show that our proposed solution outperforms the existing solutions in terms of energy consumption and task success rate. For example, it can reduce the mobile devices’ energy consumption by up to 40%.