One way latency estimation without clock synchronization

Abstract

A method for determining whether clock skew may exist between a sending node and a remote node during two-way network testing (using protocols such as TWAMP), and a computational method for revising measured latency data to compensate for clock differences. The method for compensating for clock skew comprises monitoring the network latency between two nodes during a defined time interval. When clock skew is detected, a flag is set, and, after the time interval has completed, clock skew S is estimated using the minimum latency values for the interval. The recorded latency values for the interval are then revised using the calculated clock skew S, and one-way latency results reported. The improved accuracy can be achieved with only on a few computations after the data have been collected. This a posteriori approach saves on computational resources, which can be at a premium for network testing equipment.