Reliability analysis in component-based development via probabilistic model checking
Nikola Beneš, Barbora Bühnová, Ivana Černá, Radek Ošlejšek · 2012
Engineering of highly reliable systems requires support of sophisticated design methods allowing software architects to competently decide between various design alternatives already early in the development process. Architecture-based reliability prediction provides such capability. The formalisms and analytical methods employed by existing approaches are however often limited to a single reliability measure (the probability of failure on demand) and consideration of behavioural uncertainty (focusing on the uncertainty in model parameters, not the behaviour itself). This paper presents a formal reliability assessment approach for component-based systems based on the probabilistic model checking of various reliability-related properties specified in probabilistic linear temporal logic (PLTL). The systems are formalized as Markov decision processes (MDP), which allows software architects to encode behavioural uncertainties into the models in terms of nondeterministic (scheduler-decided) choices in the MDP.