Quantitative Analysis of Software Fault-tolerance Design Modes Based on Probabilistic Model Checking

As the control core of modern systems and infrastructures, the high-reliability operation of the software is of great significance. Software fault-tolerance is one of the widely used reliability design methods, and its quantitative analysis has always been a hot topic. Based on comparing the characteristics and limitations of traditional methods, such as mathematical models, fault injection, and model-driven methods, this paper introduces probabilistic model checking (PMC) technology into the modeling and evaluation of software fault-tolerance. This method is based on probabilistic models, such as continuous-time Markov chain, modeling the system behavior of fault-tolerant design, and defining system reliability properties by temporal logic language. In particular, the Erlang model is used to realize fixed time delay for modeling those designs considering timing monitoring. This paper systematically introduces how PMC is applied to the modeling and analysis of software structure fault-tolerance, information fault-tolerance, and fault detection. The experimental results show that the system reliability of the design mode adopting error detection and correction is higher, and the system availability is higher with the design mode of watchdog and redundancy integration. This paper provides a new software fault-tolerance evaluation technology to support researchers to compare different fault-tolerance methods in the early design stage.