Data Propagation Delay Constraints in Multi-Rate Systems: Deadlines vs. Job-Level Dependencies

Abstract

Many industrial areas are faced with a continuous increase in system complexity, while systems need to satisfy stringent timing requirements, which are traditionally based on the tasks' local deadlines. However, correct functionality is subject to high-level timing requirements on data propagation through a set of semantically related tasks. Since distributed concurrent engineering is often used to deal with the complexity of such systems, violations of data propagation delay constraints are only visible at late development stages, where changes in system design become increasingly expensive. In this paper, we leverage job-level dependencies (JLDs) that can be specified at early development stages to guarantee data propagation delay constraints. Therefore, we present an approach that extends the Real-Time Systems Compiler to enforce the JLDs in actual multicore schedules. This strategy enables us to perform extensive evaluations of the effectiveness of JLDs in combination with contemporary allocation and scheduling algorithms, where we observed schedulability improvements of up to 42%. Additionally, we identified the effect of the number of available cores on the data age.