A System-oriented Approach to Efficiency and Quality of Service for Internet Service Providers

A high efficiency (i.e., the ratio between the transported traffic and the costs for transporting the traffic) and a well-defined quality of service are two very important goals for IP networks. From this follows the central question of this dissertation: How can an Internet service provider optimise the efficiency and quality of service of its network? This thesis shows that efficiency and quality of service strongly depend on the network architecture, traffic engineering, network engineering, and the interconnections. Therefore, a system-oriented approach is employed. With this approach, all previously mentioned areas are analysed and optimised, considering their mutual influences. In the context of network architectures, the focus of the research lies on methods for providing quality of service. Within this thesis, analytical models are developed in order to determine and analyse the overprovisioning factor of a plain best-effort network compared to networks that support explicit service differentiation. Furthermore, the different IETF quality of service architectures are evaluated and compared in an experimental study. In this context, a bandwidth broker for Differentiated Services networks is developed. It offers strong quality of service guarantees and a high efficiency by overbooking. Besides the network architecture, the interconnection structure (i.e., the connections to other networks) also influences the efficiency and the quality of service of a network. In this dissertation, different approaches to optimise the efficiency, reliability, and quality of service of the interconnection structure are developed and evaluated. The dissertation shows that with these approaches, significant cost savings and quality of service improvements can be realised. How the efficiency respectively the quality of service obtained with the network architecture and interconnections can be further improved by traffic engineering is also analysed within this dissertation. Weaknesses of existing approaches are identified and corrected. Due to the steadily increasing traffic volumes, capacity expansion is the most important network engineering task of a provider. Therefore, capacity expansion strategies are elaborated and evaluated in this dissertation. They also consider the influence of different network architectures and traffic engineering