Establishing Confidence and Understanding Uncertainty in Real-Time Systems
Iain Bate, David Griffin, Benjamin Lesage · 2020
A benefit with traditional static analysis approaches to single criticality hard real-time systems is that the uncertainties, and hence confidence, associated with timing requirements being met are better understood than Measurement-Based Timing Analysis (MBTA) approaches. In brief, failures are mostly accounted for by human errors or random hardware failures. With the introduction of measurement-based approaches to timing analysis to help deal with more advanced processors, the situation is much more complex. The complexity comes from new sources of epistemic failures: imperfect timing measurements from the system, approximations in the analysis, and the conscious decision that parts of the system are not always guaranteed to be scheduled in a hard real-time manner. The goal of this paper is to establish some understanding of the uncertainties based on a proposed industrial approach to MBTA and consequently how confidence in the system's timing measures can be managed. More specifically understanding the epistemic uncertainties associated with measures used for timing analysis concentrating on whether it could have been foreseen that: further testing with a given method could have avoided failures; and deficiencies with the current testing method could have been predicted.