A traffic-modeling approach to understanding the QoS correlations in network

Abstract

A detailed understanding of the mechanism of guaranteeing quality of service (QoS) in the Internet is critical both for satisfying the requirements of users and improving the performance of network. Previous studies have focused on the macroscopic features of traffic distribution and seek for the optimal solutions based on one or more targets (e.g. delay, bandwidth) to allocate network resources. A traffic-modeling approach is proposed to reveal the potential QoS correlations in network. It is claimed that our model that incorporate hard technological constraints on router and link bandwidth and connectivity, together with abstract models of user demand and network performance, can successfully presents the QoS correlations. Further experiments are taken and the results show that this sort of correlations is characterized as diffusion and fluctuation in the scale of whole network. This conclusion interprets the common phenomenon of hotspots of network load move around and clarifies the necessity to introduce the demands of users for network performance evaluation.