Preemptive Priority-based Scheduling: an Appropriate Engineering Approach 10.1.1 Calculating Response Times

Scheduling theories for xed-priority scheduling are now suuciently mature that a genuine engineering approach to the construction of hard real-time systems is possible. In this chapter we review recent advances. A exible computational model is adopted that can accommodate periodic and sporadic activities, diierent levels of criticality, process interaction and blocking, cooperative scheduling (deferred preemption), release jitter, precedence constrained processes, arbitrary deadlines, deadlines associated with speciic events (rather than the end of a task's execution), and oosets. Scheduling tests for these diierent application characteristics are described. This model can be supported by structured, object-oriented, or formal development methods. The chapter also considers the issues involved in producing safe and predictable kernels to support this computational model. 10.1 Introduction Recent developments in the analysis of xed-priority preemptive scheduling have made signiicant enhancements to the models introduced by Lui and Layland in their seminal 1973 paper 33]. These developments, taken together, now represent a body of analysis that forms the basis for an engineering approach to the design, veriication, and implementation of hard real-time systems. In this chapter we review much of this analysis in order to support the thesis that safety critical real-time systems can, and should, be built using these techniques. Preemptive priority-based scheduling prescribes a run-time environment in which tasks, with a priority attribute, are dispatched in priority order. Priorities are, essentially, static. Processes are either runnable, in which case they are held on a notional (priority-ordered) run queue; delayed, in which case they are held 222 Introduction 223 on a notional delay queue; or suspended, in which case they are awaiting an event which may be triggered externally (via an interrupt) or internally (from some other task). Most existing hard real-time systems are implemented using a static table-driven schedule (often called a cyclic executive). Priority-based scheduling has many advantages over this static approach (see Locke 35] for a detailed discussion of this issue). In essence these advantages all relate to one theme|increased ex-ibility. However, in order to challenge the role of static scheduling as the premier implementation model, priority-based scheduling must: provide the same level of predictability (of temporal behavior) allow a wide range of application characteristics to be accommodated enable dependable (safe) implementations to be supported. All of these issues are addressed in this review, which is organized as follows. The remainder of this introduction outlines a simple model of task attributes and shows how worst-case response …