Advancements in Modeling and Analysis of Multi-Processor Real-Time Systems

Nowadays, embedded real-time applications are becoming always more demanding, requiring more and more computing power. Consequently, the adoption of multi-core platforms is an unavoidable choice: nonetheless, it increases the analysis complexity due to multiple sources of unpredictability. To fully exploit the available computing power, tasks running upon a multi-core system often exhibit a parallel structure with complex internal dependencies. The analysis complexity is further exacerbated by the scheduling effects imposed by the operating systems and, sometimes, by middleware frameworks that execute the actual workload on behalf of the operating system. As a result, modern real-time systems are influenced by manifold effects, giving rise to the need for more complex scheduling models and analyses techniques. This thesis presents new models and analysis techniques for real-time workloads. The thesis is divided into three parts. The first one addresses techniques to handle dynamic real-time workloads scheduled upon multi-core platforms. The second part proposes analysis techniques for tasks modeled as direct acyclic graphs and explicitly targeting the features of specific execution platforms. Finally, the scheduling effects due to popular middleware frameworks (such as ROS and Tensorflow) are considered, proposing suitable models and analysis techniques.