An integrated bandwidth allocation and admission control framework for the support of heterogeneous real-time traffic in class-based IP networks

The support of real-time traffic in class-based IP networks requires the reservation of resources in all the links along the end-to-end paths through appropriate queuing and forwarding mechanisms. This resource allocation should be accompanied by appropriate admission control procedures in order to guarantee that newly admitted real-time traffic flows do not cause any violation to the Quality of Service (QoS) experienced by the already established real-time traffic flows. In this paper we initially aim to highlight certain issues with respect to the areas of bandwidth allocation and admission control for the support of real-time traffic in class-based IP networks. We investigate the implications of topological placement of both the bandwidth allocation and admission control schemes. We show that the performance of bandwidth allocation and admission control schemes depends highly on the location of the employed procedures with respect to the end-users requesting the services and the various network boundaries (access, metro, core, etc.). Based on our results we conclude that the strategies for applying these schemes should be location-aware, because the performance of bandwidth allocation and admission control at different points in a class-based IP network, and for the same traffic load, can be quite different and can deviate greatly from the expected performance. Through simulations we also try to provide a quantitative view of the aforementioned deviations. Taking the implications of this ‘‘location-awareness’’ into account, we subsequently present a new Measurement-based Admission Control (MBAC) scheme for real-time traffic that uses measurements of aggregate bandwidth only, without keeping the state of any per-flow information. In this scheme there is no assumption made on the nature of the traffic characteristics of the real-time traffic flows, which can be of heterogeneous nature. Through simulations we show that the admission control scheme is robust with respect to traffic heterogeneity and measurement errors. We also show that our scheme compares favorably against other admission control schemes in