Abstract: The paper studies optimal software release time problem based on G-O software reliability model, and proposes a new minimum software test cost model with VaR constraint. The analytic formula of VaR based on non-homogeneous Poisson process is obtained, and the optimal release time of software is calculated by numeric experiments based on Monte Carlo simulation. The results and sensitivity coefficients are analyzed, which shows that decision variables of the model well presents the risk preference of software project manager. The comparison of classic G-O model and VaR constrain G-O model is conducted, and the resuls show the G-O model underestimats the extreme risk of multi-faults of software after release comparing with VaR constrain G-O model.

Key words: G-O software reliability model, software optimal release, VaR constraint, minimal expect cost