McSad: A Monte Carlo-based end-to-end scheduling anomaly detection method for distributed real-time systems

• Published in: Simulation modelling practice and theory Vol. 125; p. 102746
• Main Authors: Shi, Xianchen, Zhu, Yian, Li, Lian, Li, Jiayu
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.05.2023
• Subjects: Distributed real-time systems; End-to-end timing analysis; Monte Carlo simulation; Scheduling anomaly; Distributed real-time systems; End-to-end timing analysis; Scheduling anomaly; Monte Carlo simulation
• ISSN: 1569-190X; 1878-1462
• DOI: 10.1016/j.simpat.2023.102746

A scheduling anomaly is a counter-intuitive timing behavior that seriously jeopardizes the system. Although several types of scheduling anomaly have been proposed and their effects researched, no research has been done on scheduling anomalies for end-to-end latency analysis. In this paper, we extend the concept of scheduling anomalies to end-to-end analysis and formally define these anomalies in terms of their cause–effect job chain response time. To detect end-to-end scheduling anomalies, we propose a Monte Carlo-based scheduling anomaly detection method that involves a controlled trade-off between the degree of accuracy and analysis runtime. Firstly, we create a stochastic system model using a probability distribution function instead of the WCET and propose different sampling methods for the model. Finally, we implement a system framework simulation by updating the system. The effectiveness of the proposed approach is evaluated by experiments with stochastic task models subject to different distribution functions. In addition, the experimental results show that the method has similar time-consuming for task execution time followed different distribution functions. •End-to-end scheduling anomalies are proposed and formally defined.•Constructing a stochastic distributed real-time system model.•A distributed real-time systems simulation method was established.