QoS for real-time IP traffic

Abstract

This thesis focuses on three service models, known as Best Effort, IntServ and DiffServ, and their ability of providing QoS in IP networks. The objective is to evaluate the performance of real-time traffic based on theory and testing. In addition, mechanisms that may improve the QoS performance further, like MPLS and Bandwidth Brokers, are evaluated to find a comprehensive solution for different types of IP networks. Tests are set up to focus on typical behavior of the service models, and disclose differences in their efficiency together with problems concerning fairness in the bandwidth sharing. Results obtained show the capability of IntServ and DiffServ to supply QoS guarantees for real-time traffic in contrast to Best Effort. Best Effort is fit for networks with low input load where traffic does not need service guarantees. IntServ reserves resources end-to-end, guarantees service quality and is suitable for small networks. DiffServ classifies traffic in behavior-aggregates and is very scalable, thus is fit for use in large networks. For the Internet, the results suggest that IntServ should be used in edge networks and DiffServ over MPLS in core networks.