Clock Synchronization – An Approach using Mapping Functions

Abstract:
In a distributed system most nodes maintain a local oscillator to derive time information for synchronization with other nodes. A number of clock synchronization techniques have been presented in the literature (e.g. NTP, PTP) which rely on the exchange of messages among nodes to share timing information and to adjust the offset or skew of the clocks. In this paper we present an approach which does not require any adjustments to the local clocks, but relies on achieving synchronization through clock mapping functions which map the time at one node to the time at another node. We further show how closed paths in a graph of nodes can be used to estimate the synchronization tolerance. Through experimental results using piecewise linear functions, we demonstrate the feasibility of this approach and show how clock synchronization of better than 100 ps can be achieved in Wi-Fi environments.