Dependability of iterative software : a model for evaluating the effects of input correlation

This paper deals with the dependability evaluation of software programs of iterative nature. In this work we define a new model that is able to account for both dependencies between input values of successive iterations and the effects of sequences of consecutive failures. Differently from previously proposed models, it is based on steady state prohahilities of an iteration outcome (success, benign failure or catastrophic failure) and those representing the correlation. As such it allows to analyse the effects of the correlation between successive inputs on the dcpcndahility attributes of iterative software and assumes, as starting knowledge, information usually obtainable by testing. Requiring the designers or users to determine just the steady state probabilities of an iteration outcome, rather than difficult and costly Slate transition probabilities, this model is more useful and more generally applicable.