Identification and management of inconsistencies in dynamically adaptive software systems

摘要: Computing devices now enable access to rich information about their surrounding execution environment gathered through sensor networks or system monitors. This ability allows software systems be conceived with context in mind, instead of being created isolation as traditional approaches for development. Services provided by can adapted sensed conditions rendering such services more appropriate the environment. Adaptations system's behavior take place unannounced over time. However, if not dealt carefully, adaptations could lead inconsistencies behavior. In order avoid inconsistencies, dependencies between must carefully managed so that interactions are rendered incompatible. dissertation investigates how provide guarantees predictability when it is dynamically at run Based on observation different adaptive systems, we put forward a set requirements should satisfy ensure consistency its behavioral adaptations. We propose formal basis support development consistent presence dynamic adaptations, called Petri nets. complies and particular context-oriented programming, three levels: formalization, execution, analysis. Context nets offer formalization definition them, notion Interactions formalized well-defined rules capture intention programmers high-level, while enabling low-level representation automatic verification those rules. Consistency two levels. At design-time, properties analyzed identification possible incoherence run-time satisfiability all interaction verified, hence, ensured no occur. proposed basis, tool design, manipulation, simulation interactions. work validated demonstrating usefulness analyzing existing context-aware applications, appropriateness broadening frontiers extensibility expanding itself.