Towards a Schedulability-driven Architecture Exploration for Mixed Criticality Multicore Systems

Architecture space exploration of multicore systems has been studied for more than a decade now. The main exploration factor was the workload distribution in a sense that all processing elements run fairly similar workloads, and none of the processes misses its real-time constraints. For safety critical systems, assigning the software functions to real-time tasks of the target operating system is a crucial task as it is constrained with a set of different criticality, architectural and time-related constraints. Finding a suitable architecture configuration involves many important design decisions that have a strong impact on the system safety and performance. With the increasing complexity and scale of today's systems and the large number of possible architecture configurations, identifying the efficient architectures while satisfying different (potentially conflicting) criteria becomes tedious and error-prone for designers. In this paper, we propose an automated method using multi-objective criteria to explore the architecture of multicore safety critical systems. Our method relies on a constellation of analysis techniques where we consider architectural, criticality and time requirements (schedulability). We have implemented two Matlab algorithms to perform the architecture exploration, while the final candidate configurations are analyzed for schedulability using the Uppaal model checker.