When is wavelength conversion contributing to reducing the blocking rate?

Although many studies have now appeared on solving the routing and wavelength assignment (RWA) problem with full or partial wavelength conversion, it is still unclear to which extend the addition of wavelength converters has a significant or even measurable impact on network bandwidth efficiency or on the blocking rate. Indeed, several authors have observed that the blocking rate is often unchanged even with the addition of converters depending on the traffic instances and the network topology. For this reason, some authors, e.g., Ramaswami and Sasaki (1998) or Erlebach and Stefanakos (2003) have investigated further the usefulness or the impact of adding converters. In this study, we pursue in this direction and investigate two issues: performing some preprocessing tests when solving the RWA problem with/without conversion by identifying, at the outset, some lightpaths that will be present in the optimal solution, and some of the patterns that the traffic instances must contain in order to observe an advantage of adding converters. Conclusions are that quite particular traffic patterns must be present in order for wavelength conversion to be useful, and in that case, it is sufficient to have conversion features at a very limited number of nodes