DATS: Dispersive stable task scheduling in heterogeneous fog networks

Fog computing has risen as a promising architecture for future Internet of Things, 5G and embedded artificial intelligence applications with stringent service delay requirements along the cloud to things continuum. For a typical fog network consisting of heterogeneous fog nodes (FNs) with different computing resources and communication capabilities, how to effectively schedule complex computation tasks to multiple FNs in the neighborhood to achieve minimal service delay is a fundamental challenge. To tackle this problem, a new concept named processing efficiency (PE) is first defined to incorporate computing resources and communication capacities. Further, to minimize service delay in heterogeneous fog networks, a scalable, stable, and decentralized algorithm, namely dispersive stable task scheduling (DATS), is proposed and evaluated, which consists of two key components: 1) a PE-based progressive computing resources competition and 2) a QoE-oriented synchronized task scheduling. Theoretical proofs and simulation results show that the proposed DATS algorithm can achieve effective tradeoff between computing resources and communication capabilities, thus significantly reducing service delay in heterogeneous fog networks.