TODG: Distributed Task Offloading with Delay Guarantees for Edge Computing.

Edge computing has been an efficient way to provide prompt and near-data computing services for delay-sensitive IoT applications via task offloading. However, due to the stochastic channel resources and task arrivals, it is still very challenging to design a distributed task offloading strategy for scheduling heterogeneous tasks to different edge servers with delay guarantees. In this paper, we fully exploit the joint communication and computing resource allocation for task offloading and formulate the optimal offloading problem as a non-convex stochastic optimization problem. To tackle the problem in a distributed and efficient way, we develop TODG, a joint channel allocation and task scheduling algorithm, which can achieve an adjustable trade-off between algorithm complexity and optimality. Further, we perform a comprehensive theoretical analysis on TODG, in terms of the optimality gap, delay guarantees, and impacts of system parameters. Extensive simulation results demonstrate the effectiveness and efficiency of TODG.