Learning from probabilities: Dependences within real-time systems

Realistic real-time systems experience variability and unpredictabilities, which can be compensated by potentially very pessimistic worst-cases. Recent trends apply measurement-based approaches in modeling worst-cases with a certain confidence. While observing system evolution it is possible to extract probabilistic models to the task execution with a guaranteed probabilistic version of worst-case execution time. In this work we exploit such probabilistic models in order to study the effect of dependences on the task execution time, and we apply the developed probabilistic framework to few relevant cases studies.