This paper proposes a statistical method that can be used to monitor, control, and predict the quality (measured in terms of the failure intensity) of a software system being tested. The method consists of three steps: estimation of the failure intensity (failures per unit of execution time) based on groups of failures, fitting the logarithmic Poisson model to the estimated failure intensity data, and constructing confidence limits for the failure intensity process. The proposed estimation method is validated through a simulation study. A method for predicting the additional execution time required to achieve a failure intensity objective is also discussed. A set of failure data collected from a real-time command and control system is used to demonstrate the proposed method.

A set of criteria is proposed for the comparison of software reliability models. The intention is to provide a logically organized basis for determining the superior models and for the presentation of model characteristics. It is hoped that in the future, a software manager will be able to more easily select the model most suitable for his/her requirements from among the preferred ones.

This paper deals with the problem of determining the optimum time when testing can stop and the system be considered ready for operational use. This decision, of course, depends on the model for the software failure phenomenon and the criterion used for evaluating system readiness. Using the non-homogeneous Poisson process model of Goel and Okumoto [4], two criteria are investigated, viz the software reliability and total expected cost. A simple cost model is developed and the optimum release policy, based on cost criterion, is derived. Numerical examples are used to illustrate the results.

In this paper we develop a model for the operational phase of a software system which incorporates the uncertainty of error removal and the time spent in correcting errors. Expressions for various measures of software performance, e.g. distribution of time to a specified number of remaining errors, the expected number of errors detected and corrected in time t, and software system availability are derived. Numerical examples are used to illustrate these results.