The dynamic nature of some self-adaptive software systems can result in potentially unpredictable adaptations, which may be detrimental to overall system dependability by diminishing trust in the adaptation process. This paper describes our initial work with architectural runtime configuration management in order to improve dependability and overall system usefulness by maintaining a record of reconfigurations and providing support for architectural recovery operations. Our approach---fully decoupled from self-adaptive systems themselves and the adaptation management processes governing their changes---provides for better adaptation visibility and self-adaptive process dependability. We elaborate on the vision for our overall approach, present early implementation and testing results from prototyping efforts, and discuss our future plans.
展开▼
