Using Value Dependencies to Schedule Complex Soft-real-time Applications with Precedence Constraints

Value-based approaches have been used in the real-time scheduling context, especially for so-called anytime applications. Scheduling methodologies for signal processing or similar dataflow-related applications, however, frequently rely on constraint solving techniques like linear programming and on static schemes. For soft and firm real-time applications with precedence constraints and uncertainty in the release time distribution of the individual tasks constituting the application, however, a static scheduling scheme does not suffice. We therefore suggest a value-based model to dynamically schedule anytime applications with additional real-time requirements and show how to integrate precedence constraints into this environment.