Converter placement in all-optical networks using genetic algorithms

Wavelength routed optical networks have emerged as a technology that can ef fectively utilize the enormous bandwidth of the optical fiber. Wavelength c onverters play an important role in enhancing the fiber utilization and red ucing the overall call blocking probability of the network. Placement of wa velength converters is an NP-complete problem (K.C. Lee, V.O.K. Li, IEEE Jo urnal of Lightwave Technology 11(5/6) (1993) 962-970) in an arbitrary mesh network. In this paper, new Integer Linear Program (ILP) formulations have been proposed for the static and dynamic routing and wavelength assignment problem to reduce the number of conversions. We use Genetic Algorithms (GAs ) for placing limited range wavelength converters in arbitrary mesh wavelen gth routed optical networks. The objective is to achieve near optimal place ment of full and limited range wavelength converters resulting in reduced b locking probabilities and low distortion of the optical signal. Certain heu ristics are used to obtain starting solutions for the GAs to enable it to c onverge faster. The results obtained using GAs are compared with the heuris tic method of placement. We study the range of loads for which converters a re useful. We observe that limited range converters placed near optimally a t a few nodes can provide almost the same blocking probability as full rang e wavelength converters placed at all the nodes and that increasing the num ber of converters yields only a marginal improvement in blocking probabilit y. We also observe that uniform placement of converters can be adopted in r ing networks at low offered loads. We study the effect of uniform placement at higher loads and suggest that when used with reservation in the routing algorithm, converters ensure fairness to all nodes, Simulations have been carried out on a 12-node ring network and 14-node NSFNET. (C) 2000 Elsevier Science B.V. All rights reserved.