Performance evaluation of bandwidth partitioning in broadband networks

The capability to optimally allocate network resources such as bandwidth is a key to the economical viability of modern telecommunication networks. Requirements such as, adaptive updating of bandwidth allocations to track dynamical load variations, and preferential treatment for some services at the expense of others, based on economic considerations, however, add to the complexity of the task. We propose a framework of an integrated traffic engineering system that caters to these requirements. Based on the current global network state, the system projects the future network state and determines optimal bandwidth partitioning among the competing services that maximizes network revenue while maintaining prescribed QoS (quality of service) in broadband networks. The functionality of the system within the network management infrastructure is discussed. The emphasis, however, is on highlighting the mathematical model used to evaluate the impact of various bandwidth access control policies on the call level QoS parameters such as call blocking probability, call dropping probability and throughput. Some priority based bandwidth access control policies are also defined and compared with conventional policies such as CS (complete sharing), CP (complete partitioning) and PS (partial sharing). A comparative analysis of conventional as well as priority based policies is conducted. Some simulation results are presented that reflect upon the effectiveness of priority based policies. This performance evaluation derives the engineering methods that are subsequently used to plan and dimension the networks. There are two main contributions of this work. One is the precise formulations developed to predict call level QoS parameters based on the particular policy of bandwidth partitioning. The other is the proposed, traffic measurement and engineering based, framework of the system, which is quite practical and consolidated.