KEIDS: Kubernetes-Based Energy and Interference Driven Scheduler for Industrial IoT in Edge-Cloud Ecosystem

• Published in: IEEE internet of things journal Vol. 7; no. 5; pp. 4228 - 4237
• Main Authors: Kaur, Kuljeet, Garg, Sahil, Kaddoum, Georges, Ahmed, Syed Hassan, Atiquzzaman, Mohammed
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.05.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Clean energy; Cloud computing; Containers; Data acquisition; Data processing; edge computing; Energy; Energy conservation; Energy consumption; energy minimization; Energy utilization; Environmental impact; Industrial applications; Industrial Internet of Things (IIoT); Integer programming; Interference; Internet of Things; job scheduling; Kubernetes and multiobjective optimization; Linear programming; Management; Minimization; Multiple objective analysis; Nodes; Optimization; Real time; Resource management; Task analysis; Task scheduling; Virtual environments; Weight reduction
• ISSN: 2327-4662; 2327-4662
• DOI: 10.1109/JIOT.2019.2939534

With the rapid explosion of Industrial Internet of Things (IIoT), the need for real-time data processing with enhanced flexibility and scalability has increased manifold. However, the newly evolved containerization technology offers lucrative advantages in comparison to the conventional virtual machines. However, management of these light-weight containers is a tedious task, but Google Kubernetes offers a consolidated container management and scheduling for successful execution of various lightweight containers. Nevertheless, the existing Kubernetes solutions fall short in efficiently handling the "interference" and "energy minimization" challenges in IIoT set-up. Hence, in this article, we present a competent controller, named Kubernetes-based energy and interference driven scheduler (KEIDS), for container management on edge-cloud nodes taking into account the emission of carbon footprints, interference, and energy consumption. The problem of task scheduling has been formulated using integer linear programming based on multiobjective optimization problem. In detail, KEIDS minimizes the energy utilization of edge-cloud nodes in IIoT for optimal green energy utilization. Henceforth, the applications are scheduled on the available nodes in less time with minimum interference from other applications, which in turn guarantees an optimal performance to the end-users. An extensive evaluation of the proposed KEIDS scheduler in comparison to the existing state-of-the-art schemes indicates its superior performance on real-time data acquired from Google compute cluster.