Deployment optimisation for software variants in embedded systems

This thesis introduces a framework for finding an optimal deployment of software variants to given hardware architecture. Here the family of software variants is described by feature model. A tool was developed that successfully approaches the given problem: first by selecting suitable subset of variants to be deployed individually, second by adopting genetic approach for deploying and using the Jenetics library, and third by combining the individual results based on one of three strategies proposed in this thesis. This always returns a deployment for the entire family, as long as all variants are individually eployable. Moreover, this framework gives great level of freedom to the user. It introduces an easy way to import input data in XML format or as Java objects. In addition, it provides a transparent layer for guiding the evolution process for the individual deployments, which allows the user to integrate her or his domain specific knowledge and assure better results. In conclusion, our tests showed that the expected amortised time behaves quadratically in the number of features. Efficiency, flexibility, and usability are part of the guiding lines fulfilled by this framework