Distributed Task Scheduling in Heterogeneous Fog Networks: A Matching with Externalities Method

Abstract

Fog computing shows great potential advantages over legacy cloud computing in terms of latency and efficiency. While it also poses huge challenges in task scheduling and resource allocation, which are key to reap the full benefits of fog computing. Facing the new characteristics of fog computing, distributed task scheduling and resource allocation algorithms are necessary but challenging, which will be investigated in this paper. To minimize the latency in a distributed manner, a many-to-one matching game with externalities is formulated for task scheduling problems in fog networks, considering the inter-dependency among tasks in both communication and computation. Further, a distributed algorithm called Distributed Two-sided Stable Task Scheduling (DTS2) algorithm is proposed to find a two-sided exchange-stable solution. Analytical and simulation results show that the proposed DTS2 algorithm can converge to a sub-optimal two-sided exchange-stable solution with low-complexity. Specifically, the DTS2 algorithm can greatly reduce the service delay than the one-sided stable matching. Besides, it can offer sub-optimal performance in system average delay but serve more users, compared with the optimal solution.