Performance-oriented Design Space Exploration
Anne Martens, Heiko Koziolek · 2008
Abstract—Architectural models of component-based software systems are evaluated for functional properties and/or extrafunctional properties (e.g. by doing performance predictions). However, after getting the results of the evaluations and recognising that requirements are not met, most existing approaches leave the software architect alone with finding new alternatives to her current design (e.g. by changing the selection of components, the configuration of components and containers, the sizing). We propose a novel approach to automatically generate and assess performance-improving design alternatives for componentbased software systems based on performance analyses of the software architecture. First, the design space spanned by different design options (e.g. available components, configuration options) is systematically explored using metaheuristic search techniques. Second, new architecture candidates are generated based on detecting anti-patterns in the initial architecture. Using this approach, the design of a high-quality component-based software system is eased for the software architect. First, she needs less manual effort to find good design alternatives. Second, good design alternatives can be uncovered that the software architect herself would have overlooked. I.