End-to-end quality of service prediction based on functional regression

Abstract

This work focuses on quality of service (QoS) estimation based on end-to-end active measurements. The main problem is to continuously monitor the QoS that will receive a multimedia application in a path between two agents on the Internet. We want to estimate the QoS without sending heavy multimedia tra±c in order to measure its QoS parameters (delay, losses, jitter, etc.). In this paper we extend a recent work of Ferraty et al. [5] on functional nonparametric regression. We generalize it to include a very important case on the Internet: nonstationary traffic. We apply this result to our problem describing an end-to-end active measurement methodology. We also show by simulations how it predicts the QoS of a multimedia application in a link where its capacity and buffer sizes are unknown. The cross traffic is also unknown and can be a nonstationary process. Keywords. End-to-end active measurements, nonstationary, QoS, functional regression.