Auction design for cross-edge task offloading in heterogeneous mobile edge clouds

Abstract

Task offloading is a promising technology to exploit the available resources in edge cloud efficiently. Many incentive mechanisms for offloading systems have been proposed. However, most of existing works study the centralized incentive mechanisms under the assumption that all mobile edge infrastructures are operated by a central cloud. In this paper, we aim to design the auction-based truthful incentive mechanisms for heavily loaded task offloading system in heterogeneous MECs. We first study the homogeneous MEC situation and present a global auction executed in the central cloud as a benchmark. For the heterogeneous MEC situation, we model the system as a dual auction framework, which enables the heterogeneous MECs to perform cross-edge task offloading without the participation of central servers. Specifically, we design two dual auction models: secondary auction-based model, which enables the system to offload tasks from a large-scale region in a single auction, and double auction-based model, which is suitable for the time sensitive tasks. Then the auctions for these two dual auction models are proposed. Through rigorous theoretical analysis, we demonstrate that the proposed auctions achieve desirable properties of computational efficiency, individual rationality, budget balance, truthfulness, and guaranteed approximation. The simulation results show that the secondary auction and double auction can obtain 14.5% and 4.2% more social welfare than comparison algorithm on average, respectively. In addition, the double auction has great advantage in terms of computation efficiency.