Multi-layer Network Optimization with Long Range Dependent Traffic Towards an Efficient and Green Internet

Project: Research

View graph of relations


Given the growth, complexity and size of the Internet as well as growing concerns regarding energy consumption, there is a need for new network design methodologies to achieve cost effective and energy efficient operations of the Internet. We will provide a cost-based near optimal solution for a multi-layered network serving traffic streams modeled by Long Range Dependent (LRD) processes, where a given number of transport technologies operating in layers are supported by all the switching nodes.The aim is to optimize both the routing path and the choice of transport technologies for all traffic where the cost metric includes the effect of network resource utilization in various layers. Given the size of the Internet, the solution must be scalable. Therefore we seek heuristic algorithms where traffic streams are transported on their least cost path. Since cost and routing are mutually inter-dependent, we iterate routing and capacity assignment heuristics until convergence is achieved.A novel and key feature of the project is the inclusion of LRD traffic in the optimal design problem of multi-layer networks. We use a Poisson Pareto Burst Process (PPBP), which is known to accurately model Internet traffic and exhibits large variability in flow sizes. By using a realistic stochastic traffic model, together with flow-size dependent routing as a technology which can intelligently handle this traffic, we are able to thoroughly explore an important approach for improvement in the network efficiency and performance. Both dynamically set-up paths and permanently established paths are included in our analysis.Our solutions will produce near optimal link capacity assignments for the physical and virtual links, and near optimal virtual topology. Furthermore, using realistic prediction of traffic demands and cost of energy and transport equipments, the project will provide means for predicting future technology usage, e.g., under what traffic conditions will the core Internet become all-optical.The project will also result in important fundamental contributions by providing statistical characterization of splitting and merging traffic streams modeled by PPBP processes. Additionally, we will develop solutions that will enable telecommunication engineers and researchers to design new networks, upgrade them, and predict network evolution. In this way, the project will make industry impact, potentially leading to a “greener” Internet and lowering the cost for telecommunication providers and users.


Project number9041794
Grant typeGRF
Effective start/end date1/09/1218/02/16