Improving Reliability and Latency of High Critical Tasks in Mixed Criticality Systems through Task Rescheduling

A Mixed Criticality System (MCS) consists of various hardware and software components executing tasks with different criticality levels. The criticality of a task is determined by its impact on the overall system output (e.g., safety critical and mission critical tasks or low critical and high critical tasks). In today’s world, such systems can be found almost everywhere including cars, airplanes, remotely piloted vehicles and so on. An MCS needs to be designed considering different criticality scenarios depending on the requirements of the operating environment. Significant amount of research has been dedicated to improve various parameters of MCSs such as reliability, performance, and power consumption. In this thesis, we focus on improving the reliability and execution latency of high critical tasks in mixed criticality systems running on a Hardware/Software codesign environment. The system can run in two different operating modes: low criticality mode, which is the normal operating mode of the system, and high criticality mode. We propose two different algorithms: Reliability Priority Algorithm and Latency Priority Algorithm. In reliability priority approach, the algorithm schedules all tasks in the system and returns the final reliability and the latency of the system in low criticality operating mode. In high criticality mode, the algorithm gives priority to high critical (HC) tasks over low critical (LC) ones during the scheduling process. The LC tasks are scheduled in the gaps available considering the latency constraints. The algorithm returns the overall reliability of the system and the reliability of the HC tasks in both modes of operation. In latency priority approach, we prioritize the execution latency of HC tasks over their reliability. In the low criticality mode, the algorithm schedules the tasks to the fastest components available at the point of arrival. In the high criticality mode, HC tasks are scheduled before LC tasks in order to improve the latency of HC tasks. The results of the experimental evaluation clearly show the viability of the proposed algorithms. The reliability priority algorithm increases the reliability of the HC tasks by 6.6% on the average and the latency priority algorithm improves the execution latency of the HC tasks by 23.8% on the average for the automatically generated task graphs.