Requirements- And Architecture-based Self-adaptation Of Software Systems

Increasingly, software systems are becoming complex and running in an open, changing and unpredictable environment, which makes their runtime behaviors devi-ate or even violate the requirements. Therefore, software systems are increasingly re-quired to have the self-adaptation capability in order to tackle with the complexity of software systems and the uncertainty of their running environments. Further, runtime models have become the knowledge basis of self-adaptation because models can cap-ture the knowledge of requirements and designs at an appropriate abstraction level and hide their complexities, and support query, analysis, reasoning and manipulation.Runtime models provide a runtime abstraction representation of the knowledge of the software system and its running environments. Based on such runtime models, self-adaptation mechanisms first realize model-level reconfigurations by model-based analysis and reasoning and then map to system-level reconfigurations. This thesis fo-cuses on requirements- and architecture-based self-adaptation approaches. Require-ments models capture the high-level requirements and their realizations, providing the basis for adaptation decision making. Architectural models, serving as the intermedi-ary between requirements and implementations, provide the design knowledge for the mapping between requirements reconfigurations and system reconfigurations.This thesis has four contributions. First, the thesis proposes a requirements-based self-adaptation approach to support functional and quality tradeoffs for optimizing the overall quality of software systems. Second, the thesis proposes an approach to handle the uncertainty in requirements-based self-adaptation, which limits the negative effect of uncertainty on self-adaptation and improves the effectiveness. Third, the thesis ap-plies the requirements-based self-adaptation into service-oriented architectures for the optimization of composite services. Fourth, the thesis proposes a self-adaptation ap-proach that combines requirements- and architecture-based self-adaptation to improve the effectiveness and offer more flexibility for component-based architectures. Further, we conducted detailed evaluation to validate the proposed approaches.