The Integration of Scheduling and Fault Tolerance in Real-Time Systems

Next generation, mission critical systems will require greater exibility, predictability, and reliability than is commonly found in today's systems [17]. While many advances are required to achieve the cost-e ective engineering of these systems, one important research topic is the integration of schedulability and fault tolerance. In the past this topic has been approached by engineering expensive and highly static solutions, dependent on special purpose interconnection networks and architectures and static scheduling. This approach is not suitable for most next generation systems. Two promising approaches for more dynamic situations are the use of imprecise computations and the use of planning mode schedulers [12]. What is especially attractive is that these two approaches are not competing but rather can be used in conjunction with each other. In Section 2 we rst discuss some background material on the interaction of scheduling and fault tolerance. In Section 3, we present the imprecise computation model and discuss its advantages and disadvantages. In Section 4 we present the planning mode scheduling approach with its advantages and disadvantages and suggest how planning can be combined with imprecise computation. In Section 5 we summarize key points and present a list of outstanding issues.