Osmotic Computing Platform for Smart City Applications in the Cloud-Edge Continuum

The Cloud-Edge Intelligence (CEI) paradigm advocates decentralized data preprocessing, model training, and inference on devices across the edge of the network and in the cloud. This paper proposes a containerized Cloud-Edge microservice architecture that enables model training and prediction on both Edge and Cloud nodes, allowing for flexibility and dynamic adaptation to diverse requirements and resource availability. The framework incorporates an automatic task migration mechanism, leveraging opportunistic resource management and workload distribution between the Edge and the Cloud.