| Component-based development (CBD) techniques are frequently used for the development of large-scale complex systems to reduce system development time and cost. This can be seen in many application domains, including software systems, hardware systems, and complex systems of systems. Generally in CBD, one major concern is the satisfaction of the functional as well as the Quality of Service (QoS) requirements of the system using existing components. Most of the existing works in CBD focus on functional decomposition. Methods and tools have been developed to help with the design of a system that satisfies the functional requirements. Frequently, such a system design can be realized by multiple sets of components and different compositions of the components may yield different QoS behaviors. Also, some components may be reconfigurable and can provide different QoS tradeoffs. Thus, appropriate design decisions, including component selections and parameter settings of configurable components, should be explored such that both functional and QoS requirements of the system are satisfied.; The goal of this dissertation research is to develop techniques for QoS-driven composition analysis. The approach leverages current design technologies and focuses on realizing a design with a specific composition of components for the satisfaction of system QoS requirements. A composition is a specific selection of the set of components and the specific parameter settings of the configurable components for realizing a system design. A specific composition yields specific system QoS behaviors. QoS-driven composition analysis techniques aim to model the specification of the system and component QoS aspects, explore the design alternatives (various compositions), predict system QoS properties of the design alternatives based on the component QoS properties, and make design decisions by efficiently exploring the QoS property space of the system.; The dissertation research results in a set of integrated model, techniques, and tools to carry out the QoS-driven composition analysis. The techniques and tools support automated design space exploration, evaluation, and decision making. Thus, system designers can be relieved from the burden of QoS analysis and composition decision making, while focusing on functional decomposition tasks. This is especially beneficial for the design of large-scale complex systems where the QoS property space of the system can be too large to be effectively explored manually. The research results significantly contribute to the state-of-the-art of component-based system design. |
