Improving Fault Tolerance in High-Precision Clock Synchronization

The very popular Precision Time Protocol (PTP or IEEE 1588) is widely used to synchronize distributed systems with high precision. The underlying principle is a master/slave concept based on the regular exchange of synchronization messages. This paper investigates an approach to enhance PTP with fault tolerance and to overcome the transient deterioration of synchronization accuracy during a recovery from a master failure. To this end, a concept is proposed where a group of masters negotiates a fault-tolerant agreement on the system-wide time and transparently synchronizes the associated IEEE 1588 slaves. Experimental verification on the basis of an Ethernet implementation shows that the approach is feasible and indeed improves the overall synchronization accuracy in terms of fault tolerance.