Analysis of individual flows performance for delay sensitive applications

Abstract

SLA management approaches typically adopt provisioning strategies based on aggregate traffic in order to support endtoend delay requirements of applications. They do not take into account individual flows needs in terms of delay. However, this delay can be very higher than the one observed by aggregate traffic, causing an important impact in network application performance. This paper presents a study based on simulations that makes an analysis of the endtoend delay observed by individual flows. Several scenarios are used to evaluate this performance and some metrics are proposed to investigate empirical relations that show the endtoend delay behavior when are analyzed individual flows, the aggregate traffic and the network load.