Optimizing Social Welfare for Task Offloading in Mobile Edge Computing

While mobile applications are increasing in use and complexity, the computational constraints on mobile devices remain as the bottleneck for serving computation-intensive mobile applications. Mobile edge computing (MEC) provides a computing paradigm to serve the computational demands of such mobile applications by offloading the mobile devices’ computational tasks to the edge servers. Double auction has been adopted in MEC to provide a mechanism to assign the tasks of mobile devices to the edge servers while considering the satisfaction level for both entities. We improve the double auction mechanism beyond prior research in MEC. Specifically, we construct a model to support the real-world practices in the pricing scheme of edge computing, such as that provided by Amazon, and to support the parallelizing and distributing of workloads to multiple edge servers. We propose an efficient mechanism to achieve the optimal social welfare by converting the allocation problem to a minimum cost flow problem. In addition to reaching the optimal social welfare in polynomial time computations, our proposed mechanism achieves individual rationality and strong balance budget.

I. Introduction

As mobile applications increase in their use and functionalities, the finite constraints on the mobile devices’ computations and energy become the bottleneck for such application processing. To address such issue, Mobile Edge Computing (MEC) [1] provides a capable computing paradigm to offload the computationally intensive application loads to the edge servers as opposed to processing them on the local devices. MEC provides cloud-computing capabilities and an IT service environment at the edge of the mobile networks. MEC has received increased attention in recent years and is envisioned to be a critical technique in 5G network due to its desirable properties such as low latency, proximity, high bandwidth, Specifically, mobile devices can offload some computational tasks to its adjacent edge servers to obtain a quicker response for getting the corresponding results.