A Machine-Learning-Based Auction for Resource Trading in Fog Computing

Fog computing is considered to be a key enabling technology for future networks. By broadening the cloud computing services to the network edge, fog computing can support various emerging applications such as IoT, big data, and blockchain with low latency and low bandwidth consumption cost. To achieve the full potential of fog computing, it is essential to design an incentive mechanism for fog computing service providers. Auction is a promising solution for the incentive mechanism design. However, it is challenging to design an optimal auction that maximizes the revenue for the providers while holding important properties: IR and IC. Therefore, this article introduces the design of an optimal auction based on deep learning for the resource allocation in fog computing. The proposed optimal auction is developed specifically to support blockchain applications. In particular, we first discuss resource management issues in fog computing. Second, we review economic and pricing models for resource management in fog computing. Third, we introduce fog computing and blockchain. Fourth, we present how to design the optimal auction by using deep learning for the fog resource allocation in the blockchain network. Simulation results demonstrate that the proposed scheme outperforms the baseline scheme (i.e., the greedy algorithm) in terms of revenue, and IC and IR violations. Thus, the proposed scheme can be used as a useful tool for the optimal resource allocation in general fog networks.