Improving Far-Edge Device Management in IoT Applications Using Kubernetes
Carlos Resende, João Oliveira, Filipe Sousa, Waldir Moreira, Lúıs Almeida · IEEE Open Journal of the Industrial Electronics Society · 2025
Internet of Things (IoT) driven digitalization is shifting data processing to the Edge, reducing the burden of constant Cloud communication. Advances in resource-constrained microcontroller-based IoT devices that interact with the environment, such as in Cyber-Physical Production Systems, enable them to assist in computation offloading, extending Edge Computing into the so-called Far-Edge that includes such devices. However, updating these devices often requires manual interventions, full firmware updates, or proprietary tools, leading to potential application downtime. To fully leverage Far-Edge enhanced computing capabilities, it is crucial to integrate Far-Edge devices with Cloud orchestration tools, streamlining service management and deployment along the Cloud to the Far-Edge continuum. Current approaches overlook these devices' computing power and their potential to host services, supporting IoT continuum orchestration only from the Cloud to the Edge. This paper introduces Far-edge IoT device mAnagement (FITA), the first platform that integrates Far-Edge devices into Kubernetes-based infrastructures. FITA provides a Far-Edge container- like solution compliant with the Open Container Initiative. It extends Kubernetes to support service deployment on heterogeneous Far-Edge devices seamlessly and provides a method for creating virtual representations of Far-Edge devices to expose their unique capabilities to the Kubernetes scheduler. Our evaluation shows a mean deployment time on clusters with 500 services and 100 devices of around 600 ms, and a device registration time of around 1080 ms, with CPU and memory consumption of around 23 milicores and 1500 MB, respectively. Overall, FITA improves service continuity, deployment speed, and application resilience, supporting the future of the IoT, particularly in industry.