A Taxonomy of Architectural Patterns for Self-Adaptive Systems

Autonomic systems are able to adapt themselves to unpredicted and unexpected situations. Such adaptation capabilities can reside in individual components as well as in ensembles of components. In particular, a variety of different architectural patterns can be conceived to support self-adaptation at the level both of components and of ensembles. In this paper, we propose a classication of such self-adaptation patterns { for both the component level and the system level { by means of a taxonomy organized around the locus in which the feedback loops promoting adaptation reside. We show that the proposed classication covers most self-adaptation patterns, and enables deriving further ones by applying a simple set of composition mechanisms. Three examples of existing patterns of the taxonomy are detailed in the paper to show the applicability of the approach. As discussed in the paper, the advantage of the proposed classication is twofold: it enables identifying the (possibly common) properties of the existing self-adaptation patterns; and, consequently, it can help developers in choosing the most appropriate self-adaptation patterns for the development of autonomic systems.