Abstract
The explosive growth of the Internet in the past few years is mainly attributed to the ubiquity of the IP protocol and derives from the fact that IP provides a relatively simple, unified mechanism for the intercommunication of heterogeneous networks across the Internet. IP's simplicity, though, comes at the expense of the non-existent Quality of Service (QoS) capabilities. IP supports the best-effort model and hence does not guarantee that information sent by a host will indeed reach the intended destination, nor does it provide any rate guarantees; IP will only make an earnest attempt to deliver the information. The Differentiated Services (DiffServ) QoS model was introduced as a way to provide QoS in IP networks while at the same time avoiding the complexity and scalability problems introduced in the Integrated Services model, which was conceived earlier. DiffServ achieves this by performing classification and traffic conditioning of IP flows at the network edge and forwarding the admitted aggregates through the core network. All complexity is thus pushed to the edge and the functionality within the core network remains as simple as possible. The basic DiffServ network is comprised of a number of architectural components, which reside at the edge and core routers. The mechanisms that implement these components are very important since they are responsible for determining the performance and efficiency of the DiffServ network overall. In this thesis, we propose designs and implementations for these particular components and evaluate them through simulations. In addition, we assess suitable traffic control mechanisms in terms of their capability to provide DiffServ. Asynchronous Transfer Mode (ATM) is a popular transport technology widely deployed in backbone networks. It is therefore necessary for DiffServ to be supported effectively by this technology. Approaches for mapping DiffServ to ATM service categories are examined. We propose the Guaranteed Frame Rate (GFR) service category as a potential solution for supporting the Assured Service class and we show, using simulations, that this is indeed feasible. Multi-Protocol Label Switching (MPLS) is a technology which integrates the label-swapping paradigm with network-layer routing. Its property to allow the set-up of explicit paths in connectionless IP networks makes it an attractive solution for IP traffic engineering. The inherent characteristics of MPLS also make it a very good candidate for providing Differentiated Services. Here we propose and present an architecture, which enables the support of differentiated services in MPLS-based networks.