Abstract

Software Engineering proposes several approaches for managing software production, from defining a domain to deploying the system for final use. Within those approaches are self-adaptive systems, which aim to manage variability dynamically while executing. Dynamic software product lines offer a solution by defining system variability. We present FMweb-K, a framework that improves variability management from design to execution. It integrates feature models, the MAPE-K loop, and variation points with adaptation rules linked to IoT sensors. We first validated the former through a proof of concept with a Java architecture and reconfiguration engine to manage variability using Docker. FMweb-K was validated through a proof of concept, comparing it to other solutions. Future research will employ deep learning to detect new points of runtime variation, seamlessly adding new system states.