Latency and Reliability Improvements of High Critical Tasks in Mixed Criticality Systems

Reliability and execution latency of high critical tasks are crucial for a successful execution in a mixed criticality system with tight design constraints. In this paper, we focus on two main design problems, namely latency-constrained maximum reliability problem and reliability-constrained minimum latency problem, for the applications running tasks of different criticality. The target architecture can run in two operating modes: low criticality mode (normal operating mode) and high criticality mode. For the former problem, we first find the minimum execution latency assuming the system runs in low criticality mode. Then, using this latency as lower bound, we present a heuristic algorithm to improve the reliability of high critical tasks in the application. The proposed algorithm assigns high critical tasks to the highest reliable processing elements in the technology library, and then, schedules low critical tasks without exceeding the given latency constraint. Similarly, for the latter problem, we first determine the highest reliability assuming the system runs in low criticality mode. Then, considering the overall system reliability, the proposed approach reduces the latest completion time of high critical tasks by giving them priority over low critical ones when selecting processing elements. The experimental evaluation conducted using task graphs shows up to 14.69% reliability improvement and 20.05%, on the average, latency improvement for the high critical tasks in the system.