Formalising Adaptation Patterns For Autonomic Ensembles

Autonomic behavior and self-adaptation in software can be supported by several architectural design patterns. In this paper we illustrate how some of the component- and ensemble-level adaptation patterns proposed in the literature can be rendered in SCEL, a formalism devised for modeling autonomic systems. Specifically, we present a compositional approach: first we show how a single generic component is modelled in SCEL, then we show that each pattern is rendered as the (parallel) composition of the SCEL terms corresponding to the involved components (and, possibly, to their environment). Notably, the SCEL terms corresponding to the patterns only differ from each other for the definition of the predicates identifying the targets of attribute-based communication. This enables autonomic ensembles to dynamically change the pattern in use by simply updating components' predicate definitions, as illustrated by means of a case study from the robotics domain.