Priority related QoS in TCP/IPv6 networks

Active Queue Management (AQM) algorithms have been introduced for routers supporting TCP dataflows to assist the TCP congestion control mechanism to perform satisfactorily in all circumstances. However, today the Internet is expected to provide a distinctly different service to different users based on the value and content accessed, etc. The most popular AQM algorithm, Random Early Detection (RED), is famous for its fairness-addressed characteristic, which makes Quality of Service (QoS) differentiation impossible for it, and many other RED-based algorithms, while working on their own. The introduction of the IPv6 header flow labelling ability will allow new AQM algorithms which can provide differentiated levels of QoS based on user demands to be developed. This thesis aims to develop a methodology for the AQM algorithms to work in conjunction with the IPv6 flow labelling ability, so as to provide acceptable Quality of Service (QoS) according to the end users' dynamically varying QoS requirements. We aim to offer a simple and generic process, which is adequately adoptable in Differentiated Services (DiffServ) TCP/IPv6 environments. The contribution of this thesis is four-fold. Firstly it develops a novel approach to allow AQM routers to track the QoS status of the dataflow and provide a differentiated QoS according to user requirement. Secondly, an AQM algorithm is developed from the classic Proportional Integral controller with the ability to support different QoS requirements with a faster response time. Thirdly, we have developed a mechanism to aid the AQM algorithm in predicting the queue length and improving system performance. Finally, we have proposed schemes to provide service differentiation for flows with different priorities in DiffServ Networks. We demonstrate the proposed algorithms through simulations in Network Simulator 2 (NS2). The proposed methodologies are shown to exhibit many desirable properties, such as providing differentiated throughputs and fast system response. Compared with traditional AQM techniques, they offer significant improvements in guaranteeing different levels of QoS in TCPIIPv6 networks according to user requirements and the priority of the dataflows.