Behavioural Model-based Control for Autonomic Software Components

Autonomic Managers (AMs) have been largely used to autonomously control reconfigurations within software components. This management is performed based on past monitoring events, configurations as well as behavioural programs defining the adaptation logics and invariant properties. The challenge here is to provide assurances on navigation through the configuration space, which requires taking decisions that involve predictions on possible futures of the system. This paper proposes the design of AMs based on logical discrete control approaches, where the use of behavioural models enriches the manager with a knowledge not only on events, states and past history, but also with possible future configurations. We define a Domain Specific Language, named Ctrl-F, which provides high-level constructs to describe behavioural programs in the context of software components. The formal definition of Ctrl-F is given by translation to Finite State Automata, which allow for the exploration of behavioural programs by verification or Discrete Controller Synthesis, automatically generating a controller enforcing correct behaviours. We implement an AM by integrating the result of Ctrl-F compilation and validate it with an adaptation scenario over Znn.com, a self-adaptive case study.