Unreliable distributed timing scrutinizer: adapting asynchronous algorithms to the environment

We present an unreliable distributed timing scrutinizer (UDTS). This mechanism captures the state of the network and, based on its observation, estimates the waiting time that maximizes the efficiency of a round-based protocol regarding the number and/or the duration of its computational rounds. An UDTS is not an abstract oracle, it is implementable and reveals the properties of the network. We illustrate our approach by addressing the consensus problem in asynchronous distributed systems equipped with an unreliable distributed timing scrutinizer. The proposed protocol is conceptually simple. It benefits from the UDTS predictions to implement different convergence and decision conditions. This allows the protocol to be efficient in number and/or duration of rounds for stable networks.