Real-time fault-tolerant atomic broadcast

摘要

We present algorithms for real-time fault-tolerant uniform atomic broadcast. We first design a distributed execution model for asynchronous systems with crash failure (called synchronized phase system (SPS)), then we give an algorithm for atomic broadcast in SPS. In an SPS, the processes try to run in synchronized sounds like in synchronous systems. SPSs can be implemented in asynchronous systems, but the liveness properties follow the properties of the knowledge of processes concerning the failures of other processes. In timed partially synchronous systems, we can give explicit feasibility conditions to solve real-time uniform atomic broadcast. At present, these algorithms are being implemented in the French project ATR.