A State Machine-Based Approach for Reliable Adaptive Distributed Systems

Adaptive systems are often composed of distributed components that co-operate in order to achieve a global behaviour, and yet many approaches for adaptive systems are centralised or make strong assumptions about the distributed aspects of the problem. However, if insufficient attention is paid to the problem of decentralisation, especially in the difficult and unpredictable environments in which adaptive systems are commonly deployed, it can introduce inefficiencies, and even cause catastrophic failure. An adaptive system is either required to implement subtle synchronisation and consensus protocols or accept certain types of failure from which the system cannot recover. A major goal of our research is to facilitate the development of adaptive, reliable and distributed applications. We provide a framework in which a state machine language is used to define logically centralised behaviour. This is automatically translated into a reliable and efficient distributed implementation that enforces the correct co-ordination in the presence of unpredictable failures.