Elements of Self-adaptive Systems – A Decentralized Architectural Perspective

Software is evolving towards a greater complexity and variability, with a continously changing environment. In this context, self-adaptive systems are acquiring a great relevance. Their architectures are inherently dynamic and by definition, also reflective. However, their self-referential nature might compromise their compositionality, and even the use of the architectural approach. This work intends to decide on its suitability, by considering its fundamentals in detail. After some initial definitions, the nature of both self-adaptation and self-organization is discussed, and the implicit modular structure is determined. Then a tentative taxonomy of elements in self-adaptive architectures is provided, which is also discussed in a decentralized setting. To support our initial hypothesis about the suitability of architectures, the algebraic properties of their composition is studied in detail. Then, the suitability of a reflective approach in this context is considered, and then a concrete example of an autonomic system is described, using a reflective architectural description language. The chapter concludes discussing the suitability of this approach, and how the architectural perspective of self-adaptation does not actually imply a centralized topology.