Estimating reliability for OSS: An approach with change-point in operational phase

Abstract

Software reliability growth models (SRGMs) proposed on the foundations of Non-Homogenous Poisson Process (NHPP) have been the most accepted way to analyze reliability growth of a software since 1970's. Henceforth, Literature witnesses the evolution of SRGMs in terms of different assumptions, parameters, performance etc. This study is an attempt to model reliability growth phenomenon for Open Source Software (OSS). As the development process followed by OSS significantly differs from proprietary software, traditional software engineering SDLC models can't be followed for its development. OSS development procedure does not incorporate a dedicated testing phase and therefore negligible effort expenditure is done on testing of software prior its release. The volume of volunteer participation in its operational phase thus becomes a significant criterion in the representation of its reliability growth. Also, the impact of sudden changes in fault detection process (change-point) due to several reasons like variations in fault density, irregular team size, volunteer expertise etc is captured in proposed SRGM. This paper suggests a reliability model based on user growth with the concept of change point to better cater the heterogeneity in software failure process. Failure data of GNOME 2.0 is used in this paper reliability growth analysis. Unknown parameters are estimated using Least Square Estimation regression technique. Performance inspection of proposed SRGM is carried out using goodness-of-fit criteria like Mean square error (MSE), Coefficient of Determination (R2), Predictive Ratio Risk (PRR), and Predictive Power (PP). The study recommends that introduction of the change-point in SRGM improves prediction capability of the model and hence establishes an evidence of its applicability.