Analysis of Input/Output Paradigms for Real-Time Systems

摘要

The correctness of a real-time system with hard deadline requirements dependsboth on the logical correctness and on the timing correctness of the system. The principles of rate monotonic scheduling have proven to be very useful in providing a framework for designing, analyzing, and modifying the timing and concurrency aspects of real-time systems. This paper illustrates how to build a mathematical model of the schedulability of a real-time system, taking into consideration such factors as preemption, synchronization, non-preemptibility, interrupts, and process idle time. In particular, this paper illustrates how these principles can be applied to input/output interfaces (e.g., to devices or local area networks) to predict the timing of various design alternatives. The primary characteristic that distinguishes real-time system from non-real-time systems is the importance of time. The correctness of a real-time system depends not only upon its logical correctness but also its timing correctness. System complexity tends to compromise correctness unless there are techniques and methods for managing the complexity. Basic software engineering principles such as abstraction, encapsulation, and information hiding from the basis of methods and techniques that are used to manage logical complexity. Rate monotonic scheduling theory offers a set of engineering principles for managing timing complexity. (KR)