Database Pointers: A Predictable Way of Manipulating Hot Data in Hard Real-Time Systems

Abstract

Traditionally, control systems use ad hoc techniques such as shared internal data structures, to store control data. However, due to the increasing data volume in control systems, these internal data structures become increasingly difficult to maintain. A real-time database management system can provide an efficient and uniform way to structure and access data. However the drawback with database management systems is the overhead added when accessing data. In this paper we introduce a new concept called database pointers, which provides fast and deterministic accesses to data in hard real-time database management systems compared to traditional database management systems. The concept is especially beneficial for hard real-time control systems where many control tasks each use few data elements at high frequencies. Database pointers can co-reside with a relational data model, and any updates made from the database pointer interface are immediately visible from the relational view. We show the efficiency with our approach by comparing it to tuple identifiers and relational processing.

This work is supported by ARTES, a network for real-time research and graduate education in Sweden.