Predicting the Number of Defects Remaining In Operational Software
Patrick Hartman · Naval Engineers Journal · 2001
ABSTRACT Software is becoming increasingly critical to the Fleet as more and more commercial off‐the‐shelf (COTS) programs are being introduced in operating systems and applications. Program managers need to specify, contract, and manage the development and integration of software for warfare systems, condition based monitoring, propulsion control, parts requisitions, and shipboard administration. The intention here is to describe the state‐of‐the‐art in Software Reliability Engineering (SRE) and defect prediction for commercial and military programs. The information presented here is based on data from the commercial software industry and shipboard program development. The strengths and weaknesses of four failure models are compared using these cases. The Logarithmic Poisson Execution Time (LPET) model best fits the data and satisfied the fundamental principles of reliability theory. The paper presents the procedures for defining software failures, tracking defects, and making spreadsheet predictions of the defects still remaining in the software after it has been deployed. Rules‐of‐thumb for the number of defects in commercial software and the relative expense required to fix these errors are provided for perspective.