Incorporating faults and fault-tolerance into real-time networks: a graph-transformational approach

Abstract

The introduction of fault tolerance into real-time systems presents particular challenges because of the price of redundancy and the added complexity of verification and validation on these redundant structures. This thesis brings structural and formal design techniques to bear on this problem. Verification of fault tolerance properties in such systems has only received limited attention. in particular the design methodologies are in their infancy. We propose a transformational design methodology, specific to a real-time systems architecture. We then reason about the compositional addition of fault tolerant components and templates of the derived designs. This requires that we show the existing axiomatic semantics for our chosen architecture sound with respect to a more constructive semantic model. The issues of presenting an operational model for a real-time architecture are discussed and a model is proposed. The extension of the existing semantics, to allow for faulty behaviour, is shown to preserve the existing semantic properties and the application of our methodology shown to be usable by a sizeable study. The contribution of this thesis is to define a transformational design methodology in which components can be extracted from a design and replaced by another component preserving functionality while providing fault tolerance. This approach requires the precise modelling of the faults we consider. the transformational method and verification of the transformed design with respect to faults.