Proving the safety of hard real-time systems can be quite challenging, even for comparatively simple systems. Therefore, many different analysis and verification techniques exist to ensure the safety of such systems. Usually, these techniques rely on a pure mathematical methodology, which comes with the benefit that they are mathematically proven. However, the problem is that they abstract from reality and therefore neglect external parameters like execution overhead, cache effects, and interrupts. In this paper, we examine how the execution overhead of a Real-Time Operating System (RTOS) can impair the timing of preemptive tasks under fixed priority scheduling. Based on this examination, we develop an extended version of the wellknown fixed-point iteration for response time analysis to consider said overhead. The results being calculated with this extended version are then compared to the results calculated by the original analysis. We further compare these results to actual measurements of the response time from real system implementations, which are done by utilizing non-intrusive hardware tracing.
展开▼
