Abstrakt
Feedback control loops that monitor and adapt managed parts of a software system are considered crucial for realizing self-adaptation in software systems. The MAPE-K (Monitor-Analyze-Plan-Execute over a shared Knowledge) autonomic control loop is the most influential reference control model for self-adaptive systems.
The design of complex distributed self-adaptive systems having decentralized adaptation control by multiple interacting MAPE components is among the major challenges. In particular, formal methods for designing and assuring the functional correctness of the decentralized adaptation logic are highly demanded.
This article presents a framework for formal modeling and analyzing self-adaptive systems. We contribute with a formalism, called self-adaptive Abstract State Machines, that exploits the concept of multiagent Abstract State Machines to specify distributed and decentralized adaptation control in terms of MAPE-K control loops, also possible instances of MAPE patterns.
We support validation and verification techniques for discovering unexpected interfering MAPE-K loops, and for assuring correctness of MAPE components interaction when performing adaptation.