Abstract

In recent years, there have been many studies on research questions related to reducing ICT energy consumption. The aim is to save energy, improving network efficiency while at the same time to meet user perceived quality of service requirements. This presentation will report on a new multilayered network optimization approach where traffic between each origin-destination (OD) pair follows Poisson Pareto burst process. This is an important step beyond previous studies that assumed deterministic traffic demands because it enables the study of flow size dependent routing. Accordingly, we classify traffic according to size of flows with the purpose to find the best route and technology/layer for each traffic stream (characterized by its OD and flow size range) and best capacity assignments on all virtual and physical links such that the overall network cost is minimized while meeting required grade of service. Given that the optimization is non-linear and given that the solution must be scalable to realistic size networks, we use a heuristics based on shortest path in each layer and fixed point iterations. We demonstrate that even for large networks, near optimal solutions is achievable in order of minutes. Validation of the solutions will also be discussed. To manage the vast amount of data associated with large networks, we use the Network Markup Language (Netml) which provides also a friendly user interface. The interface and the optimization have the potential to lead to a valuable network design tool for industry. Finally, a video that demonstrates the optimization and the user interface will be shown.