Integrating fault-tolerant and real-time requirements of distributed systems

The authors propose a distributed dynamic action scheme that allows a recovery concept permitting efficient distributed computing during normal operation to be combined with efficient exception handling in the case of an effective error. They provide a dynamic action model tailored to the dynamic nature of distributed processing. This model offers a recovery concept which allows the recovery region of a recovery line to surmount the size of the corresponding computation in order to gain high efficiency during normal operation. By running the versions of a recovery block as distributed actions, it becomes possible to incorporate a recovery concept that allows efficient distributed processing during normal operation and prompt reaction in the case of error by running the different versions in parallel. To implement the dynamic action model efficiently a redundant recovery graph keeps track of recovery regions. On the basis of this graph the authors provide decentralized protocols that produce a consistent system state that is fast, efficient, and concurrent with normal system activity