Abstract
In today's competitive market, service providers seek to increase their revenues while keeping capital and operational expenditures down. This entails the provision of new revenue-generating multimedia services and guaranteed bandwidth for business- and mission-critical applications. It also means a requirement to consolidate disparate networks into a single infrastructure, eliminating the need to maintain several physical networks. With various networking technologies around, integrating them into one seamless infrastructure is a big challenge. The main theme of this thesis is to investigate this issue from the viewpoint of a service provider. Designing such an infrastructure would require an understanding of the properties of the traffic generated by these services in order to determine the network resources required. The issues of traffic modelling investigated leads to this thesis developing a new multi-layer modelling approach. This work provides insight into future cross-layer optimisation approach. These emerging services also require varying performance guarantees from the network. Hence the network must be QoS-aware. For IP-based networks, IntServ and DiffServ are the two QoS architectures proposed. This thesis discusses these approaches and proposes an integrated architecture to achieve a scalable end-to-end connectivity. On the broadband front, ATM is widely deployed in the core/backbone of high-speed networks for its support for QoS for real-time services. MPLS was initially developed as a new transport platform promising the robustness of IP routing at connection-oriented switching speeds. It has met this goal and currently is finding new found importance in IP traffic engineering, an important tool in achieving the QoS, performance guarantees and operational efficiency. This thesis proposes the addition of both MPLS and ATM to the end-to-end architectural framework for the new generation IP QoS. It also looks at how IPv6 can play an important role towards the evolution of this architecture.