Lighting Fibers in a Dark Network

In the 1990's with the advent of wavelength division multiplexing, considerable experimental and theoretical work was undertaken in the study of network design inherent to these transparent or clear channel optical networks. One class of traffic engineering models in this setting are the {em routing and wavelength assignment} problems. These are in contrast to traditional networks where traffic flow paths need not be assigned an end-to-end wavelength or frequency. The additional requirement means that the cost of optimal capacity allocations for "WDM-flow" can in general be arbitrarily worse than for their traditional counterparts. We refer generally to the increased network design cost as the transparency gap. Considerable research has addressed the tradeoffs between various network architectures built from different switching functionalities. Counterbalancing bad news that the transparency gap may be arbitrarily large in general, are studies showing that the increase in cost may be modest under restrictive conditions such as the presence of a limited number of wavelength converters, suitable network topologies, or partial wavelength switching (e.g., spatial switching only) in multi-fiber networks. We explore reasons for why this is the case especially for dedicated (or static) multi- wavelength multi-fiber transparent networks. For a number of reasons, we focus specifically on the network design problem and particularly on the cost incurred as traffic on the network grows. For this setting, we give a simple theoretical confirmation that supports previous findings which show that the transparency gap all but disappears in multi-fiber networks with wavelength selective routing, and to some degree for waveband routing.