Con-Pi: A Distributed Container-based Edge and Fog Computing Framework for Raspberry Pis

Edge and Fog computing paradigms overcome the limitations of Cloud-centric execution for different latency-sensitive Internet of Things (IoT) applications by offering computing resources closer to the data sources. In both paradigms, single-board small computers like Raspberry Pis (RPis) are widely used as the computing nodes. RPis are usually equipped with processors having moderate speed and provide supports for peripheral interfacing and networking. These features make RPis well-suited to deal with IoT-driven operations such as data sensing, analysis and actuation. However, RPis are constrained in facilitating multi-tenancy and resource sharing. The management of RPi-based computing and peripheral resources through centralized entities further degrades their performance and service quality significantly. To address these issues, a framework, named \textit{Con-Pi} is proposed in this work. It exploits the concept of containerization and harnesses Docker containers to run IoT applications as microservices on RPis. Moreover, Con-Pi operates in a distributed manner across multiple RPis and enables them to share resources. The software system of the proposed framework also provides a scope to integrate different application, resource and energy management policies for Edge and Fog computing. The performance of the proposed framework is compared with the state-of-the-art frameworks by means of real-world experiments. The experimental results evident that Con-Pi outperforms others in enhancing response time and managing energy usage and computing resources through distributed offloading. Additionally, we have developed a pest bird deterrent system using Con-Pi to demonstrate its suitability in developing practical solutions for various IoT-enabled use cases including smart agriculture.