Developing Fault Tolerant Distributed Systems by Refinement

Distributed systems are usually large and complex systems composed of various components. System components are subject to various errors. These failures often require error recovery to be conducted at architectural-level. However, due to complexity of distributed systems, specifying fault tolerance mechanisms at architectural level is complex and error prone. In this paper, we propose a formal approach to specifying components and architectures of fault tolerant distributed and reactive systems. Our approach is based on refinement in the action system formalism - a framework for formal model-driven development of distributed systems. We demonstrate how to specify and refine fault tolerant components and complex distributed systems composed of them. The proposed approach provides designers with a systematic method for developing distributed fault tolerant systems.