Reliability certification of software components

Reuse is pinpointed as a key factor to improve productivity and reliability of software systems. Verification and validation of software components and the resulting system is important for reuse to be beneficial on a broad industrial basis. This paper suggests a modelling approach which is suitable for reliability certification of modular systems. It discusses a general reliability certification procedure and provides guidelines and opportunities for how to certify software components and also presents some alternatives for certification of modular systems. We conclude that to create reliable systems, we must start certifying the individual constituents of the systems. 1: Introduction Reuse, modularization, and certification of software are three areas given a lot of attention in the software engineering community today [1]. This paper focuses on the importance of combining these three areas. Object-oriented techniques, or at least modular techniques, are pinpointed as being suitable for reuse, but it must be remembered that components often are not reused if their reliability cannot be guaranteed. Therefore it is essential to realize that reliability certification is a must when discussing reuse. We use the term certification to distinguish reliability certification from verification and validation in general. The term component is used throughout this paper as a generic entity which can be certified and reused. Components to be retrieved from a repository must have a quality stamp in terms of what level of reliability can be expected from them as they are put into a system. The reliability must reflect the intended usage of the component, i.e. a component may be viewed as being reliable for one user and unreliable for another depending on the intended usage of the component. This paper focuses on the certification of software reliability for a modular system. Reuse in general and object orientation in particular are discussed frequently, but few of these discussions highlight the problem of actually assigning a reliability measure to the components being put into the repository. This paper focuses on techniques available for certifying both components as well as systems. The certification process includes two major activities: usage specification (consisting of a usage model and usage profiles) and certification procedure using a reliability model. This paper provides insight into the reliability certification domain to support others wishing to evaluate and validate the models and methods presented in this paper. At this stage of the research, no validation project has been run, but we hope that …