A practical degradation model for mixed-criticality systems

Existing Mixed Criticality System (MCS) task models consider criticality as the relative importance of a task and use it to achieve graceful degradation of the system either by suspending or degrading tasks with relatively lower criticality level than the overloading task. Graceful degradation based on this notion of criticality as a relative parameter may not always be desired. In this paper, we propose a Context-Aware Mixed Criticality System (CA-MCS) model with which each task can be specified with multiple degraded budgets. To handle a system overload due to timing faults i.e. budget overrun of a task, tasks that need to be degraded are chosen irrespective of their criticality level. However, to handle system overload when multiple tasks overrun their budgets, a specific degraded budget of the task chosen for degradation is decided based on the criticality level of the overrun tasks. Experiments performed in a realistic automotive testbed confirm the benefit of CA-MCS model when compared with the state-of-the art degradation strategies. Results show that contextaware degradation gives the ability to degrade the performance in a controlled manner by effectively isolating the effects of degradation between applications. Finally, we derive a sufficient schedulability test under fixed priority scheduling scheme for the CA-MCS model.