Design of Multi-layer Telecommunication Networks: Fairness, Resilience, and Load Balancing

Abstract: Migration to Next Generation Internet architectures poses new challenges for network operators in planning core networks and calls for efficient network planning and optimization tools. Optimization models
underlying such tools are developed in this thesis. We study a number of single and two-layer core network design problems defined as mathematical programmes, focusing on fair bandwidth allocation among demands, recovery mechanisms, and load balancing on network links.

Assuming elastic traffic, fair allocation of network bandwidth among the users is not trivial since different users may have different preferences and requirements for minimum bandwidth. We study single and two layer network dimensioning tasks where elastic and non-elastic
demands are combined, and investigate different fairness principles, with special attention devoted to proportional fairness. The models are developed for designing the networks for the normal state of network
operation, as well as for failure states.

For the two-layer problems it is not at all clear in which layer the recovery should be performed, and what recovery mechanisms to use. Therefore, recovery aspects in different layers are studied and models are provided for different recovery mechanisms. Furthermore, a generic resolution framework and heuristic algorithms for the selected
dimensioning and allocation problems in two-layer networks are developed.

Balancing of load on network links decreases probability of rejection of future requests due to shortage of resources in some parts of the network. In the thesis different load balancing options are discussed, and an integrated routing, recovery, and load balancing strategy is developed. It combines failure dependent backup path protection,
shortest path routing, and load balancing according to proportional fairness principle.

The thesis presents both theoretical findings, models, and resolution algorithms for the studied problems. Efficiency of the algorithms is illustrated by numerical examples. The thesis also gives a systematic view and classification of different aspects related to network architecture, recovery, fairness, and flow/congestion control.