Over the last years, the number of IoT devices has grown exponentially. Thus, Fog and Edge computing move part of the computation closer to data sources, exploiting interconnected devices in a computing continuum viewpoint. These devices are heterogeneous in terms of performance, features, and capabilities, requiring proper programming models and run-time management layers. This work presents a first version of the BarMan open-source framework. We developed a run-time task allocation policy to maximize application performance and through an experimental evaluation performed on a real cluster, we evaluated different execution scenarios. The results show an improvement up to 66% on the frame processing latency with respect to a monolithic solution.

Measurement-based WCET methods are reliable as long as we know all the application behaviors at design-time. In particular, we need to observe all the possible execution time behaviors before the actual system run-time to get a reliable estimation. This is, unfortunately, difficult to achieve. In this work, we propose an online monitoring framework with the goal to detect, at run-time, unexpected application behaviors and trigger the probabilistic-WCET re-estimation when needed. The online monitoring and estimation framework allows dealing with unexpected situations, such as never-seen applications, faults, or incorrect offline estimations. The proposed approach is tested with simulated and real data.