Software reliability demonstration testing with consideration of damage size of software failures

Abstract

This article describes an extension of the conventional reliability demonstration testing method that is based only on the number of software failures. A reliability demonstration testing method that also considers the size of the damage by the failure is proposed. Both the software used continuously in time (continuous model) and the software used discretely in time (discrete model) are considered. For each model of reliability demonstration testing, the producer's risk and the consumer's risk are formulated. A method is shown by which the testing can be designed when those risks are specified. Solution procedures are shown for the case where the damage size follows an exponential distribution and for the case where it follows a Poisson distribution. The characteristics of the proposed method are discussed using numerical examples. © 1999 Scripta Technica, Electron Comm Jpn Pt 3, 82(5): 10–21, 1999