Models for the Consistent Interaction of Adaptations in Self-Adaptive Systems

摘要

Self-adaptive systems enable the run-time modification, or dynamic adaptation, of a software system in order to offer the most appropriate behavior of the system according to its context of execution and the situations of its surrounding environment. Depending on the situations currently at hand, multiple and varied adaptations may affect the original behavior of a software system simultaneously. This may lead to accidental behavioral inconsistencies if not all possible interactions with other adaptations were anticipated. The behavioral inconsistencies problem becomes even more acute if adaptations are unknown beforehand, for example, when new adaptations are incorporated to the system on the fly. Self-adaptive systems must therefore provide a means to arbitrate interactions between adaptions at run time, to ensure that there will be no inconsistencies in the system's behavior as adaptations are dynamically composed into or withdrawn from the system. This chapter presents existing approaches that allow the development of self-adaptive systems and management of the behavioral inconsistencies that may appear due to the interaction of adaptations at run time. The approaches are classified into four categories: formal, architectural modeling, rule-based, and transition system approaches. Each of these approaches is evaluated with respect to the assurances they provide for the run-time consistency of the system, in the light of dynamic behavior adaptations.