Fault-tolerant clock synchronization for distributed systems using continuous synchronization messages

Abstract

We present a probabilistic synchronization algorithm which sends periodic synchronization messages, instead of periodic bursts of synchronization messages as other algorithms do. Our "continuous" approach therefore avoids the burst network loads of other algorithms. Nodes always have current estimates of other nodes' clocks, allowing them to monitor the state of system synchronization, and adjust their clocks as needed. The algorithm is fault-tolerant, and may be easily adapted to a wide variety of systems and networks. We analyze and simulate the algorithm's performance on a 64-node hypercube, and show that the algorithm provides tight synchronization while imposing only a light load on the network.<>