On the schedulability of a data-centric real-time distribution middleware

This work presents an analysis of the Data Distribution Service for Real-Time Systems (DDS), a data-centric distribution middleware that supports the development of predictable applications, from the schedulability point of view. The study focuses on how DDS intends to guarantee the real-time behavior through the mechanisms included in the standard, and proposes some extensions to this standard. Furthermore, the paper looks at other approaches to build distributed systems based on object distribution and remote procedures calls which can guarantee predictability, and shows how to use DDS to obtain real-time applications. A set of concepts defined in the Modelling and Analysis of Real-Time and Embedded systems (MARTE) standard has been integrated into DDS in order to allow using Model-Driven Engineering (MDE) and schedulability analysis techniques. Finally, to emphasize the results obtained from the analysis, the paper also performs a brief evaluation to validate the predictability of a particular DDS implementation.

► We analyze the “Data Distribution Service for Real-Time Systems” (DDS) standard and its capabilities for real-time systems.
► Mechanisms included in the specification are not enough to provide hard real-time guarantees.
► DDS enables the schedulability analysis and provides a smooth transition from the MARTE model to the real system.
► We provide simulative analysis to evaluate the predictability and overhead.
► We conclude that multiple real-time aspects are not included in the standard but open to implementations.