This Thesis addresses the problem of resource management in ATM (Asynchronous Transfer Mode) and IP (Internet Protocol) networks with QoS (Quality of Service) support. ATM and IP technologies allow, nowadays, the integration of different types of services in the same network infrastructure. However, the variety of services and their requirements, involve big challenges at the resource management level, which needs to be as efficient as possible.

This Thesis is divided in two parts. The first part of the Thesis addresses the resource management of ATM and IP access networks with QoS support. It is presented an historical perspective of the access networks evolution. In terms of ATM access networks, resource management strategies based on VPs (Virtual Paths) are proposed, and dimensioning methodologies are defined, which take into account the QoS requirements at the call and cell level. The considered resource management strategies allow the establishment of trade-offs between the resource utilization and the signaling load. In terms of IP access networks, it is proposed a new network architecture that represents an evolution of the legacy access networks, allowing a larger resource sharing. This architecture allows QoS differentiation and support for multimedia applications. More specifically, it is proposed a solution that includes the integrated support of recently introduced technologies to establish and configure multimedia sessions, manage the QoS policies including AAA (Authentication, Authorization and Accounting) functions, and reserve network resources.

In the second part of the Thesis, two scalable call admission control mechanisms are addressed: probing mechanisms and mechanisms based on the aggregation of individual reservations. Both mechanisms allow resource management without the maintenance of the per-flow state in each network element. The probing mechanism estimates the network QoS level, through the insertion of probe flows, to decide if a new flow can be accepted or not. A new probing mechanism is proposed, denoted by e-probing, which minimizes the resource stealing problem that exists in the simple probing mechanism when it is applied to a system with multiple service classes. Analytical models are developed and simulation studies are performed to analyze the resource stealing problem and to determine the influencing factors on the estimation of the loss ratio performed by the probing and e-probing flows. The e-probing mechanism obtains, simultaneously, a high resource utilization and service differentiation without resource stealing. The results obtained with this mechanism were also validated through the development of a probing test-bed. In the mechanism based on the aggregation of individual reservations, the core network elements only need to maintain the state of the flows’ aggregate (instead of individual flows), and the aggregates’ bandwidth is dynamically adjusted. Analytical models are defined and simulation studies are performed to analyze the trade-off between signaling load and resource utilization. These studies show that the hierarquization of the network, that is, its partition in smaller areas, and the configuration of aggregates between area border routers, as opposed to end-to-end aggregates between domain border routers, reaches an utilization close to the one of per-flow signaling with a signaling load significantly smaller.