Model-based adaptivity in real-time scheduling

This paper presents an approach for building schedulers for real-time, especially soft real-time systems. It is based on two technologies, model-based computing, and self-adaptivity. The scheduler operation is based on on-line evolving models that reflect all significant information on the running system. We argue that despite the obvious overhead added by these computationally rather expensive technologies, these costs are returned as a result of more intelligent scheduling. A typical soft-real time device, an autonomous network probe device is used as a proof of concept application.