Abstract

Passive Optical Networks (PONs) have evolved to provide much higher bandwidth in access networks. They involve low cost equipment, fiber infrastructure, maintenance, and appear to be one of the best candidates for a major component of the next-generation access networks. A PON is capable of delivering bundled voice, data, video and other multimedia services. It is a point-to-multipoint optical network, where an optical line terminal, located at the central office, is connected to multiple ONUs through some intermediate nodes equipped with either optical splitters or array waveguide gratings. In this thesis, we have focused on PONs with splitters only and have developed different optimization models for multicasting and wavelength assignment, i.e., for efficient provisioning of PONs. We consider two different objectives: Maximizing the weighted number of granted requests and maximizing the weighted number of served ONUs under the condition of a limited number of wavelengths. Variants include combinations of the two objectives in order to capture a cost/revenue concern through appropriate definitions of the weights. We developed different models dealing with either a tree or a light mesh topology as well as with traffic-grooming, i.e., Time Division Multiple Access ( TDMA) or without traffic-grooming. Models correspond to integer linear programs and for some of them, we have favored a column generation modeling in order to solve the linear programming relaxation and obtain a scalable solution process. Simulations have been conducted to validate all the models (with/with traffic-grooming). We have also evaluated the performance of the traffic-grooming models to observe the grade of service in terms of the number of served ONUs, bandwidth, bandwidth usage/waste, as well as efficient provisioning of PONs, under three different traffic scenarios for the objective of maximizing the weighted number of ONU s.