Topology Design and Capex Estimation for Passive Optical Networks Attila MITCSENKOV, Géza PAKSY, Tibor CINKLER Department Of Telecommunications and Media Informatics Budapest University of Technology and Economics Budapest, Hungary {mitcsenkov|paksy|cinkler}@tmit.bme.hu Abstract—Several optical access network technologies are available for network operators providing broadband services (FTTx: Fiber-to-the-X solutions). These technologies are now in deployment phase, therefore network and topology design issues play an increasingly important role. In this paper we address broadband optical access network design minimizing deployment costs, taking operation issues into account, using detailed cost and network models of the above listed FTTx technologies that suit best to actual networks due to detailed cost metrics used instead of just minimizing fiber lengths. We present a heuristic solution that works fast even for large problem instances, providing results with a difference less than approximately 10-20% from the computed ILP (Integer Linear Programming) optimum for smaller cases where ILP could be used. Along with these algorithms we present case studies of real-life network and service requirement instances (number of customers ranging from 400 to 20.000). Keywords – Passive Optical Network; Network deployment; Topology planning; Optical access network; Broadband access; PON; FTTx; CAPEX I. INTRODUCTION Access network technologies being deployed in the near future have to face high bandwidth requirements and growing needs for real-time traffic or reliability. Network operators can decrease administrative, maintenance and management expenses by offering triple-play service in a single network, giving the potential to replace the separate networks for voice, data and video traffic, and use a single converged network with common control plane and management resources. Promising technologies fulfilling these requirements use optics even in the last mile access, as close to the customers as possible, e.g. passive optical networks (PON), active Ethernet or VDSL networks. These are referred to as Fiber-to-the-X (X stands for Home, Building, etc.) network architectures [1]. These are mature and standardized technologies, being deployed currently or in the near future. However, details of topology and implementation often determine success and profitability of a given network technology. Therefore, theory has to be put into practice, and high performance topology optimization methods are needed to ensure low deployment costs as well as working on actual geography (map) data and service requirements [2]. In this paper we address Passive Optical Network (PON) topology planning, paying regard to deployment cost minimization, along with trivial operational aspects. The presented algorithm uses geography data or existing access network topologies as input. Results are obtained within seconds or a few minutes, depending on the network size even for 10.000s of customers. The presented heuristic algorithm provides topologies with an overall cost of approximately 10% over the optimal solution achieved by an Integer Linear Program (ILP) – at least for smaller problem instances where ILP works. A. Related work Several papers have been published addressing technical issues, e.g. traffic distribution or upstream fiber access mechanisms for Passive Optical Networks; however topology design and network deployment had lower interest. Several solutions still exist in the literature for the PON network planning problem with different performance and speed characteristics. A polynomial time 2-approximation algorithm was presented in [5], however it offers fast operation in the expense of rough approximation of the optimum. In [6] and [7] the authors investigated Simulated Annealing, Tabu Search and evolutionary algorithms for fiber-VDSL network planning. The problem addressed in [8] is slightly similar to that presented in this paper, based on a different and slightly simpler cost function. It gives an ILP optimal solution and a heuristic approach for PON network planning, providing 10% higher cost than the optimum. Finally [9] compares a Genetic Algorithm-based semi-automatic optimization tool with networks designed manually, reporting approximately 8% gap between them. Comparison of these methods is somewhat difficult, due to the different cost functions and reference methods used. For that very reason, the cost function used in this paper is designed to reflect total practical deployment costs, including the equipments and the cable plant as well. As a reference, an ILP-based lower bound was used, being really close to the optimum. The work described in this paper was carried out with the support of the BONE-project ("Building the Future Optical Network in Europe”), a Network of Excellence funded by the European Commission through the 7th ICT- Framework Programme