Re-Certification of Software Reliability without Re-Testing

Abstract

Usage testing or operational profile testing is depicted as an important test technique to remove the most critical faults from an operational perspective. The technique also allows for determination of the software reliability through application of software reliability models. This is, of course, beneficial, but a problem arises as the usage changes and this can be expected as new services are added to the existing system or the behaviour of the users change due to some reason. Therefore, a method to re-certify software reliability without re-testing the software for all potential changes that may occur is needed. This paper outlines such a procedure based on fault content estimations and by recording a number of measures during usage testing. The procedure gives a basis for determining whether more testing or other means to remove faults is needed prior to the usage changes or if the reliability requirements are fulfilled even after a change in usage. It is concluded that the proposed procedure can be a valuable tool to cope with changes in usage, but more research as well as practical experience from the proposed procedure is needed.