Architectures for run-time adaptation

Since several years, a major trend in industrial development has been to shift development-time variability to runtime. As computing power is increasing and cost of memory and disk space is decreasing, delaying decisions until runtime has become a common solution to increase flexibility in many application areas. For this reason, adaptive systems have emerged as a vibrant research area comprising multiple disciplines. Multiple approaches for developing adaptive systems have been published that focus on different aspects. However, most approaches deal with adaptation on an abstract model level and it is still unclear how these model adaptations can be translated to concrete mechanisms that can be used for code generation or manual implementation. We intent to address this gap by researching available mechanisms, design patterns and styles that can be used to realize different adaptation types. For this reason, we first identify a set of typical adaptation scenarios and discuss different realization approaches. We then define a reference architecture for adaptive systems that captures the necessary requirements for the adaptation logic and executing system that are necessary in order to apply these mechanisms. Additionally, we implement a prototype system in order to validate our reference architecture and to provide practical experience on the realization of adaptive systems.