Distributed Task Offloading in Cooperative Mobile Edge Computing Networks

Mobile edge computing (MEC) has been advocated as a promising technique to handle computation-intensive and delay-sensitive mobile tasks at edge servers. However, the disparate computing resource distribution among edge servers and the high communication and computation cost make it very challenging to find an efficient task offloading decision. This paper considers a task offloading problem with constraints over a cooperative MEC network, where multiple edge servers and a cloud center collaboratively process tasks received from mobile devices. We formulate such a cooperative task offloading problem as a general convex constrained optimization problem. Different from existing models that focus on a single performance measure, our problem formulation can represent diverse performance metrics, such as service latency, energy consumption, and a combination of both. Moreover, in addition to standard linear constraints, we consider nonlinear inequality constraints owing to resource limitations or delay requirements. To solve the formulated convex optimization problem, we develop a novel distributed algorithm based on gradient projection and virtual queue techniques. The proposed distributed algorithm enables each edge server to compute its task offloading decision via local communications only, which does not require any central processors with possible communication/computation bottlenecks. We show that the proposed algorithm converges to the optimal workload allocation decision at a sublinear rate. Simulations verify the efficiency of the proposed algorithm and demonstrate that it has lower computational complexity than the baseline algorithms.