Towards Decentralized Self-adaptive Component-based Systems

摘要

Modern software systems are challenging challenge software engineers since they must adapt effectively and efficiently with respect to the environment in which they are deployed. To this end, the paper outlines an architecture, supported by special-purpose languages and aspect oriented techniques, for the design of component-based of distributed self-adaptive systems. Software artifacts are dynamically grouped, and constantly supervised by a network of ad-hoc components. The supervision mechanism exploits special-purpose languages that define data collection, correlation, aggregation, and analysis, to reason on properties defined at different levels of pervasiveness (from component-wide to system-wide). Its goal is to identify situations that trigger adaptations. Our approach is characterized by a decentralized architecture in which there is no single point of failure or bottleneck, and a clear separation of concerns between business logic and adaptation mechanisms. These concepts are demonstrated by a fantasy example that offers a general abstraction for distributed load balancing problems.