Software reliability estimations/projections, cumulative & instantaneous

Musa's methods (J.D. Musa et al., 1987) for software development and test planning combined with Duane's learning curve approach (E.O. Codier, 1968) for hardware reliability growth testing provide an efficient means for estimating and demonstrating reliability requirements. This paper reviews the analyses of Tractenberg (M. Trachtenberg, 1985) and Downs (T.Downs, 1985) that provide a foundation for Musa's basic (linear) model. I have used Musa's basic model in combination with an approach similar to that used by Duane and Codier for derivation of a formula for instantaneous failure rate for hardware to develop formulas for the estimation of instantaneous failure rate for software. These calculations show significant correlation with interval estimates and provide an efficient method for showing the achievement of goal reliability for software without a separate demonstration test.