Resiliency of distributed clock synchronization networks

Abstract

Clock synchronization refers to techniques and protocols used to maintain mutually consistent time-of-day clocks in a coordinated network of computers. A (clock) synchronization network is an interconnection of computers to implement a particular clock synchronization solution. To prevent clock-dependency loops, most synchronization networks use a stratified approach which is essentially a tree structure with a Primary Reference Clock (at "stratum- 0"). A node at stratum-i+1 exchanges synchronization messages with its parent node at stratum-i and also with some other nodes at same or other level. The purpose of this redundancy is two fold: (i) to calculate smoother steering rate adjustment, (ii) to maintain connectivity in the event of a failure. We provide an analytical framework to evaluate the performance of different approaches for resilient synchronization networks. To evaluate resiliency of synchronization networks, we characterize failure recovery metrics like connectivity and failure detection delay in terms of parameters related to network topology and failure recovery solutions.