Timing and Timing Control

Abstract

A real-time computer system must provide its results at specified instants of physical time. But what is time? After a short look at the history of time, this contribution investigates time measurement in a distributed computer system and introduces the notions of precision and accuracy of a global time base. It then discusses different methods for the synchronization of the local clocks of a distributed computer system. Since the global time base of a safety-critical real-time computer system must not fail if any one of its constituent physical clock fails, the topic of fault-tolerant clock synchronization is treated in detail. The final section elaborates on the benefits that can be accrued from the existence of a fault-tolerant global time base in a large real-time system. One of the most important benefits is the achievement of temporal predictability and the simplification of the interfaces among the nodes of a large distributed computer system by the use of time-triggered protocols. Time-triggered protocols provide error detection without failure propagation from a faulty receiver to a correct sender.