Project B21

Optical Access Networks

DFG Research Center Matheon Technische Universität Berlin
Duration: October 2009 -
Project director: Prof. Dr. Martin Skutella, Dr. Andreas Bley
Institut für Mathematik, Technische Universität Berlin, Straße des 17. Juni 136, 10623 Berlin, Germany.
Tel: +49 (0)30 - 314 78654 / +49 (0)30 - 314 78846
email: {skutella,bley} 'at'
Researcher: TBA
Support: DFG Research Center "Mathematics for Key Technologies" (Matheon)

Project description.

Background and Motivation.

Modern telecommunication networks are based on optical technology that admits to route data traffic at extraordinarily high bit-rates. Network operators in Germany plan to provide optical network access for every household as quickly as possible. This aim has also found its way to politics such that several months ago the enthusiastic goal was set to connect every household to the optical backbone network before the year 2018.

This project aims at the development of mathematical models and efficient solution algorithms to support cost-effective long term evolution planning for optical access networks. The major challenges in this context are the complicated technical restrictions caused by the new optical access technologies, the huge dimension of regional and nation-wide access networks, and the necessity to migrate existing networks to the novel technology smoothly over several years.

In contrast to backbone networks, that usually have higher connectivity to ensure survivability in case of node or edge failures, access networks typically feature a tree-like structure in order to keep costs small. Novel optical access technologies support these structures via innovative channel splitting devices and low cost passive optical elements. On the other hand, these technologies impose various side constraints, such as combined bounds on the lengths of paths and the node degrees in these tree networks. Related basic mathematical optimization problems that take some of these constraints into account have been studied in the literature. None of these models, however, takes the actual constraints imposed by optical access technologies fully into account.

The enormous size of their networks forces regional and nation-wide access network operators to migrate to the new technologies step by step over several periods. To cope with the practical requirements that arise when migrating existing networks to new technologies, more flexible models are needed, which take both the multi-period planning aspect and the variability and uncertainty in the demands into account. Also, novel algorithmic methods must be developed in order to compute evolution plans for networks of practically relevant size.

Research program.

The aim of this project is to develop mathematical optimization tools that support network operators in the design and deployment of optical access networks. For the construction of such a network, the operator has to make many decisions, the most important of which are:

The active optical nodes play an important role in access networks. They provide the last piece of ``smart'' technology when routing a signal from the backbone network to the user. An active node can balance out attenuation and dispersion effects stemming from signal splitting devices and long optical fibers.

In this context, it is important to place active nodes within the optical network cleverly. Here, we find interesting and complicated combinations of facility location and routing problems with different and to some extent conflicting restrictions: On the one hand, the objective to keep construction costs low requires to choose only few locations for the expensive active node technology. On the other hand, the chosen active nodes still have to guarantee high quality signal routing from every user to the optical backbone network.

New mathematical challenges arise from innovative capacity sharing concepts that permit the connection of multiple users to the network via a single optical transmission channel. This novel technology is the key for connecting many users to the optical network cost-efficiently. However, sharing channels causes additional disturbances from attenuation or dispersion. The complicated interplay between the number of users connected via a shared channel to an active node, the distances of these users from the active node, and the bandwidth available for each of these users implies new restrictions to the location and routing problem that have not arisen in network planning so far.

With the described decisions about the network structure, the planning process, however, is not over. The sheer size of the considered networks makes it impossible to migrate a regional or nation-wide access network towards optical technology all at once. Mathematically, we face a multi-period network planning problem, where we have to decide when to built which part of the network such as to maximize the overall profit. The (partially optical) network has to be functioning at every point in time and the possible network migration and reconfiguration actions are budgeted. As an integrated problem we must deal with complicated time dependent issues in the network structure when routing user demands.

Another important aspect in this context is that technologies and economic conditions may change over the planning horizon. To cope with this uncertainty, we have to find a robust network evolution concept that permits to rearrange construction plans over time such that the latest side conditions can be integrated into an optimal solution without revoking completed steps. This makes the time dependent network evolution already an important and challenging topic on its own.

The complexity of optical access networks exceeds that of existing networks by far. Therefore, it is an extra challenge to find mathematical tools that do not only support the general planning process for a network, but also in particular the planning process for optical access networks of realistic size. Existing solution approaches already fail when it comes to solving simpler problems of realistic, yet comparatively small size. The goal of this project is to understand the particular structure of optical access networks, to find appropriate mathematical models, and to develop mathematical tools that can deal with the newly encountered network size and the challenges of the optical access network technology at the same time.


Within Matheon