next up previous
Next: Fault-Tolerance! Up: Towards Fault-Tolerance Aspects for Previous: Towards Fault-Tolerance Aspects for

Real-Time distributed embedded systems

Two fundamentally different paradigms for the design of real-time systems are event-triggered and time-triggered architectures. In an event-triggered architecture, all activities - task activation, communication, and so on - are initiated as consequences of events (significant state changes). Because event-triggered architectures make all scheduling and communication decisions on line, they are sometimes called dynamic architectures or interrupt-driven architectures. Time-triggered architectures, on the other hand, are driven by the progression of the global time. All tasks and communication actions are periodic, and external state variables are sampled at predefined points in time. Time-triggered architectures are based on stronger regularity assumptions than event-triggered architectures and are therefore less flexible but easier to analyze and test. If the real-time system is time-triggered, then, it is known how to solve problems of replica determinism, systematic testing for timeliness, and timely membership service. The Time-Triggered Protocol (TTP) [KG94] is an integrated communication protocol for time-triggered architectures. It provides the services required for the implementation of a fault-tolerant real-time system: predictable message transmission, message acknowledgment in group communication, clock synchronization, membership, rapid mode changes, and redundancy management. It implements these services without extra messages and with only a small overhead in the message size.


next up previous
Next: Fault-Tolerance! Up: Towards Fault-Tolerance Aspects for Previous: Towards Fault-Tolerance Aspects for
Tolga Ayav 2004-10-25