Optimization algorithms for data center location problem in Elastic Optical Networks

Modern optical networks must meet many challenges following from the quick and permanent development of IT industry. The latest trend that focuses much attention is cloud computing - a concept that enables delivery of various IT services over the Internet. From the network perspective, the cloud computing implies a growth of the traffic volume and advent of new traffic patterns. In this work, we combine the cloud computing topic with a new concept of optical networks, namely, Elastic Optical Network (EON), which allows for efficient and scalable bandwidth provisioning. In more detail, we address the problem of data center location with lightpath provisioning in EONs. We focus on optimization algorithms for locating data centers with the objective to minimize the amount of spectrum required to serve anycast demands. We propose effective heuristic algorithms that utilize the well-known optimization technique called Column Generation. We present results of numerical experiments run on a real topology of an European backbone network. Our heuristics perform well comparing to optimal results provided by the CPLEX solver. Moreover, we report the increase of spectrum requirement in years 2012-2020 according to traffic demand patterns generated according to Cisco forecasts.