A Perspective on the State of Distributed Real-Time Research and Practice


E. Douglas Jensen

Distributed real-time research, development, and application, are all advancing with deplorable slowness. One reason is that "distributed system" is popularly interpreted very broadly, while "real-time" is popularly (and incorrectly) interpreted very narrowly. The intersection of these two interpretations is consequently quite small. Moreover, these small intersections are predominately in currently fashionable application areas having very limited functional and non-functional properties with respect to either, much less both, distribution and real-time. Hence very few significant scientific and engineering challenges are being addressed by most users, vendors, and even many researchers. As I define a distributed real-time computer system, its one core characteristic is that the system includes trans-node behaviors (the nature of which depend on the system's programming model), and the predictability of the end-to-end timeliness of those behaviors must be acceptable to the application under the current conditions. This encompasses a wide range of extant (not to mention prospective) systems, although they are not usually perceived as such, and that interferes with advancing their science, technology, and use. This is illustrated by examples from products and practice, and from contemporary research.