A model and algorithm for sizing and routing DCS switched telecommunications networks

Abstract

Demand for broadband services such as fax, videotelephony, video conference and data transmission continues to explode as we move into the twenty-first century. The new broadband demand differs from voice traffic in that it varies rapidly with respect to the average length of time capacity is held by a customer. Hence, steady state models of network traffic are not valid in general, and may not provide approximations that are sufficiently accurate for network design. In addition, modern telecommunications networks incorporate advanced switching technology that can provide flexible routing of network traffic based on network load and projected demand. It is desireable to take advantage of this new flexibility to design reliable, yet low cost, networks. In this dissertation a multistage stochastic linear programming model for the design of broadband networks is presented, along with a specialized algorithm for solving the program. The algorithm is based on Network Recourse Decomposition (NRD) first introduced by Powell and Cheung. The solution method incorporates cost calculations that prove to be useful for both sizing and routing decisions.