Real-time scheduling — Distributed scheduling

The complexity of real-time processes to be controlled or supervised, the high number of data and events to be processed, the topological distribution of processes, on the one hand, and the emergence over the last few years of networks dedicated to embedded systems, on the other, are all factors that have led to a rethinking of centralized real-time applications. Today, the notion of distributed architecture is widely used in industry, particularly in the transportation sector. By way of example, the application domains in which distributed architectures are most commonly used are :

• control of transport equipment (avionics, automotive, railways, etc.);

• urban traffic control and regulation ;

• industries (process control) ;

• military applications (missile trajectory tracking, etc.);

• aerospace (satellite tracking, automatic piloting, etc.);

• home automation (home security...) ;

• telecommunications (switching systems, etc.) ;

• the medical field (assistance and monitoring of patients, etc.).

A computer system designed to control or supervise operations is usually made up of several processing units (computers, programmable logic controllers, ECUs [Electronic Control Units], etc.), sensors, actuators, peripherals for visualization and dialogue with operators. All these elements are interconnected by a network or by several interconnected networks (industrial networks, office networks, field buses, etc.). This type of system is known as a distributed real-time system.

In this type of system, the scheduling of tasks and messages plays a fundamental role in their temporal validation. In this article, we present the different scheduling analysis techniques for tasks, depending on their type, and for messages, depending on the type of network used (CAN, AFDX...).