Bounding the maximum sampling rate when measuring PLP in a packet buffer

Abstract

Network Sampling is vital for estimating network parameters. Recent seminal work has proposed that there are theoretically derived upper limits to sampling rate. In this paper we explore the maximum sampling rate to use for estimating packet loss probability given a fixed number of samples, by simulation. Our results show that there is a strong relationship between the sampling rate and the accuracy of measurement. We find the maximum sampling rate for optimal measurement of PLP for specific network scenarios given a fixed number of samples. An important area of application of this work is certain Measurement Based Admission Control scenarios, where faster measurement is important.