Defining Quantiles for Estimating IP Packet Delay Variation in NGN Core Networks

Abstract

Traffic delay is one of the important metrics used for evaluating network performance. Delay and delay variation characteristics of IP packets transferred over multi-section networks can be derived, estimated or composed from component distributions of IP package delay in each network section. Approximate methods are needed in the cases of unknown or complicated delay distribution functions, which are unavailable or unusable in practice. The ITU-T has proposed a method for estimating IP packet delay variation. One of noticeable factors affecting the estimation accuracy is the packet delay population quantile which has not been adequately considered. The objective of this paper is to examine the optimal range of quantiles used for estimating the IP packet delay variation in the NGN (Next Generation Network) core networks. The paper is composed from the following ideas. Firstly, several concepts and mathematical formulas related to delay metrics based on probability and statistics theory are defined. The approximate method of ITU-T for estimating the IP packet delay variation in a multi-section network is revised. Then, another method based on convolution for composing the empirical IPTD distribution functions is proposed for the same target as the first one. Secondly, a number of test cases are implemented to measure the IP packet delay on several sections of an NGN core network. Sample data are used for computing and estimating the IP packet delay variation for multi-section networks by two methods with certain hypotheses. Finally, these methods are compared and evaluated both theoretically and empirically in regards to the estimation accuracy versus quantiles of the IP packet transfer delay. The best range of quantiles is determined to ensure the accuracy of the estimation method applied for the NGN core network.