Reliability Growth Modeling for OSS: A Method Combining the Bass Model and Imperfect Debugging

Abstract

In this highly competitive era of technology, for software to sustain in the market, it has to maintain its quality. Software reliability is one of the metrics to determine software quality. As very few efforts are spent on testing open source software, the reliability of open source software hugely depends on the number of users working on it after release. This study proposes new non-homogeneous Poisson process-based software reliability growth models incorporating factor of user growth in reliability growth of open source software. To represent user growth phenomenon in the proposed SRGMs, the Bass diffusion and Kenney’s models are used. The models are proposed for scenarios of both imperfect debugging and perfect debugging. Reliability analysis is carried out on real-world failure dataset (GNOME 2.0), and a parallel comparison among all SRGMs on four goodness-of-fit criteria (mean square error, coefficient of determination, predictive ratio risk, and predictive power) is performed. It is observed that SRGMs which are considered imperfect debugging outperforms its perfect counterpart which is consistent with realistic situations.