Sensitivity of Software System Reliability to Usage Profile Changes

Usage profiles are an important factor in software system reliability estimation. To assess the sensitivity of a system's reliability to changes in the usage profile, a Markov based system model is used. With the help of this model, the statistical sensitivity to many independent changes can be estimated. The theory supports both absolute and relative changes and can be used for systems with or without a terminal state. With this approach it is possible to very quickly estimate the uncertainty on the predicted reliability calculated from a Markov model based upon the uncertainty on the usage profile. Finally the theory is applied to an example to illustrate its use and to show its validity.