Elementary versus composite interfaces in distributed real-time systems

To manage the complexity of large distributed real-time systems, a system must be partitioned into nearly autonomous subsystems that communicate across small well-specified interfaces. The form and placement of these interfaces determines the system structure and its understandability. This paper investigates two different types of unidirectional information-flow interfaces, the elementary interface and the composite interface. It is argued that systems that interact by elementary interfaces support composability, and are easier to understand and validate than systems that are based on composite interfaces.