Scheduling Real-Time IoT Workflows in a Fog Computing Environment Utilizing Cloud Resources with Data-Aware Elasticity

Abstract

In a fog computing environment the supplementary public cloud resources should be managed as effectively as possible, utilizing a dynamic scaling mechanism, in order to provide monetary cost savings and resilience against workload fluctuations. Furthermore, the dynamic scaling strategy should take into account the data dependencies of the workload, in order to prevent data loss. Towards this direction, in this paper we propose a reactive data-aware dynamic scaling mechanism for the provision of cloud resources, along with a heuristic for the scheduling of real-time Internet of Things (IoT) workflows, in a three-tier IoT-fog-cloud architecture. The performance of the proposed scheme is evaluated and compared via simulation to a static provisioning case, under different patterns of incoming workload. The simulation results provide useful insights into how each workload pattern affects the performance of the framework under study, in each of the provisioning cases of the cloud resources.