This dissertation investigates the capacity required by an internet application in tandem with the network connection type (dial-up or broadband). An internet user’s experience in accessing various types of applications with either high bandwidth or low bandwidth is examined in a consumer choice model of broadband adoption. A consumer implicitly values the time-saving benefits derived from a higher speed internet connection used to access a particular internet application, and compares those utility benefits to the higher price of high speed connection services in making the decision to shift to broadband or remain with a dial-up connection. We find that using broadband rather than dial-up to run bandwidth intensive applications presents considerable gains in the implicit value of time saved. Assuming that internet users are rational utility maximizing agents, a logit model is used to calculate the likelihood of broadband adoption as a function primarily of the nature/type of the internet applications (“elastic or inelastic”). While the empirical results are generally consistent with our hypothesis that consumers are more likely to subscribe to broadband if they regularly run applications that are bandwidth intensive, the results vary somewhat by model specification, and are potentially sensitive to controlling for endogeneity. Correcting for endogeneity remains the major challenge in extending this research. Research Question: What is the relationship between consumer valuation of the net benefits of using bandwidth intensive applications and the adoption of broadband internet? Hypothesis: The less a consumer requires bandwidth intensive applications; the lower is the likelihood of switching from a low level bandwidth internet service like dial-up to a high level bandwidth internet type like broadband. While this relationship may appear obvious, it has not been systematically investigated or measured, nor has its importance in affecting lags in broadband adoption been adequately appreciated.

This research proposes a new dynamic energy management framework for a backbone Internet Protocol over Dense Wavelength Division Multiplexing (IP over DWDM) network. Maintaining the logical IP-layer topology is a key constraint of our architecture whilst saving energy by infrastructure sleeping and virtual router migration. The traffic demand in a Tier 2/3 network typically has a regular diurnal pattern based on people‟s activities, which is high in working hours and much lighter during hours associated with sleep. When the traffic demand is light, virtual router instances can be consolidated to a smaller set of physical platforms and the unneeded physical platforms can be put to sleep to save energy. As the traffic demand increases the sleeping physical platforms can be re-awoken in order to host virtual router instances and so maintain quality of service. Since the IP-layer topology remains unchanged throughout virtual router migration in our framework, there is no network disruption or discontinuities when the physical platforms enter or leave hibernation. However, this migration places extra demands on the optical layer as additional connections are needed to preserve the logical IP-layer topology whilst forwarding traffic to the new virtual router location. Consequently, dynamic optical connection management is needed for the new framework. Two important issues are considered in the framework, i.e. when to trigger the virtual router migration and where to move virtual router instances to? For the first issue, a reactive mechanism is used to trigger the virtual router migration by monitoring the network state. Then, a new evolutionary-based algorithm called VRM_MOEA is proposed for solving the destination physical platform selection problem, which chooses the appropriate location of virtual router instances as traffic demand varies. A novel hybrid simulation platform is developed to measure the performance of new framework, which is able to capture the functionality of the optical layer, the IP layer data-path and the IP/optical control plane. Simulation results show that the performance of network energy saving depends on many factors, such as network topology, quiet and busy thresholds, and traffic load; however, savings of around 30% are possible with typical medium-sized network topologies.