Adaptive Estimation of Clock Skew and Different Types of Delay in the Internet Network

Abstract

The end-to-end delay is often used to analyze network performance. There are different types of delay in the network: (artificial) delay due to unsynchronized clocks, transmission and propagation delays, and delay jitter. First, we show how to obtain a least-squares estimate of the clock skew (i.e. difference between the sender and receiver clock frequencies) and the fixed delay. We then show that the linear programming algorithm proposed elsewhere to estimate the clock ratio is equivalent to the maximum likelihood estimator if the delay jitter is modeled as an independent, identically distributed random process with an exponential probability density function. Finally, we show how to estimate the delay jitter and propose an unbiased recursive least-squares algorithm to estimate the clock skew and fixed delay.