Low latencies connections and decentralized servers are currently showcasing a new potential for distributed computing. By moving away from traditional centralized cloud models and toward edge computing, which allows for more autonomy and decision-making at the network's edge, almost any physical thing can be turned into an Internet of Things (IoT) device that can elaborate on data it senses from its environment. In this context, service management and adaptation routines in a highly dynamic and geographically distributed federation depends on a large number of factors ranging from performance to cost and the fluctuation of the data quality.

This paper presents mechanisms for monitoring resources at the edge in real-time, orchestrating service provisioning, performing data-driven decisions on behalf of applications, adapting service locations, and coordinating sensing tasks. The demonstration focuses on autoscaling of containers, service placement and distributed sensing, while considering utility metrics to help achieve a fluid workload in Kubernetes clusters.