Uncertainty Analysis of Race Conditions in Real-Time Systems

Race conditions in real-time systems may cause unexpected computing result. Due to the uncertainty of realtime systems, a race condition detected by many static and dynamic approaches may occur in one execution environment but may not occur in another execution environment. In this paper, an easy and practical approach based on probabilistic models is presented to analyze the uncertainties of race conditions of real-time systems in various execution environments. The approach adopts a probabilistic occurrence within a time interval to represent the uncertainty of event occurrences. The confidence level is defined to measure the accuracy of the time interval observed, and then the uncertainties of event orders are analyzed according to the relations of time intervals. We propose a T-matrix to describe the uncertainties of event orders, and a metric is presented to measure the uncertainties of executions of real-time systems in various environments. Moreover, another metric is introduced to measure the total risk of the real-time system caused by race conditions in various environments.