Resilient Estimation of Synchronization Uncertainty through Software Clocks

Many pervasive and distributed systems require clocks synchronized to a unique and external reference time. Unfortunately the actual difference between the time of a local clock and the reference time, called offset, is a variable factor that is very hard to compute: synchronization mechanisms typically estimate the offset, without offering guarantees of closeness of the estimate to the true value. The paper gives a comprehensive description of the Reliable and Self-Aware Clock (R&SAClock), a newsoftware clock that provides both the current time and the synchronization uncertainty, that is a self-adaptive and conservative estimate of the offset, thus being able of monitoring synchronization quality and detecting clock failures. Two different solutions are presented for the evaluation of the synchronization uncertainty which differ in terms of performance and coverage, and are evaluated in the context of distributed wearable mobile devices intended for railway trackside workers.