Adaptable Fault Tolerance for Real-Time Systems

This paper proposes a framework for software-implemented, adaptive fault tolerance in a real-time context. It extends previous work in two main ways: by including features that explicitly address the realtime constraints; and by a flexible and adaptable control strategy for managing redundancy within application software modules. This redundancy-management design is introduced as an intermediate level between the system design (which may itself consist of multiple levels of design) and the low-level, non-redundant application code. Application designers can specify fault tolerance strategies independently for the individual application modules, including adaptive strategies that take into account available resources, deadlines and observed faults. They can use appropriate design notations to notify the scheduling mechanisms about the relative importance of tasks, their timing requirements and both their worst-case and actual usage of resources. Run-time efficiency can thus be improved while preserving a high degree of predictability of execution.