On the impact of fast failure detectors on real-time fault-tolerant systems

Abstract

We investigate whether fast failure detectors can be useful—and if so by how much—in the design of real-time fault-tolerant systems. Specifically, we show how fast failure detectors can speed up consensus and fault-tolerant broadcasts, by providing fast algorithms and deriving some matching lower bounds, for synchronous systems with crashes. These results show that a fast failure detector service (implemented using specialized hardware or expedited message delivery) can be an important tool in the design of real-time mission-critical systems.