Model-based development of fault-tolerant real-time systems

The design of fault-tolerant real-time systems is a complex task. The system must not only satisfy real-time requirements, but it must also deliver the specified functionality in the presence of both hardware and software faults. To achieve fault-tolerance, the system has to use redundancy. This redundancy is usually achieved by replicating hardware units and executing the application within a distributed system. Model-based design tools promise to reduce the complexity of the design process by raising the abstraction level. However, most of the existing tools focus only on functional aspects. Code realizing non-functional requirements such as fault-tolerance mechanisms, communication, and scheduling is not targeted. However, this type of code makes up the majority of the code of a fault-tolerant real-time system. This work presents a model-based development tool for the design of faulttolerant real-time systems. The tool focuses on the code generation of nonfunctional requirements and therefore complements existing tools. The major contribution of this thesis is the presentation of adequate models that can be used to model fault-tolerant systems and generate the code automatically. These models comprise a formal description of the hardware architecture, the software components and their temporal behavior, the fault assumptions, and the selected fault-tolerance mechanisms. Using a template-based code generator, the fault-tolerant real-time system is generated. The template-based code generator allows an easy expansion of the code generation functionality and therefore offers a solution to handle the heterogeneity of fault-tolerant systems. The thesis presents a generic architecture for fault-tolerant systems and discusses the realization of well-known faulttolerance mechanisms in this context. Finally, the thesis outlines how formal methods can be integrated to prove the correctness of the generated code. Two complementary applications are used to demonstrate the practicability of the approach. One application points out that control times in the range of a few milliseconds can be achieved using standard hardware. The second application demonstrates that by combining different tools, one major goal in software engineering can be achieved: the development of a complex and distributed embedded system in a complete model-based way.