Genetic algorithm quality of service design in resilient dense wavelength division multiplexing optical networks

• Published in: IET communications Vol. 2; no. 4; pp. 505 - 513
• Main Authors: KAVIAN, Y. S, RASHVAND, H. F, REN, W, NADERI, M, LEESON, M. S, HINES, E. L
• Format: Journal Article
• Language: English
• Published: Stevenage Institution of Engineering and Technology 01.04.2008; John Wiley & Sons, Inc
• Subjects: Applied sciences; Computer networks; Exact sciences and technology; Flexibility; Genetic algorithms; Information, signal and communications theory; Multiplexing; Networks; Operation, maintenance, reliability; Optical communication; Optical fiber communications; Optical telecommunications; Optimization; Organization and planning of networks (techniques and equipments); Quality of service architectures; Signal and communications theory; Systems, networks and services of telecommunications; Telecommunications; Telecommunications and information theory; Wavelength division multiplexing; Europe; Performance evaluation; Mobile radiocommunication; Backbone; Optical telecommunication; Wireless telecommunication; Frequency allocation; Network architecture; Wavelength division multiplexing; Routing; Optimization; Resilience; Topological structure; Optical fiber network; Wide area network; Simulation; Genetic algorithm; Coding; Telecommunication network; Optical fiber communication; Delay time; Planning; Personal communication networks; Service quality
• DOI: 10.1049/iet-com:20070312
• ISSN: 1751-8628

The important role of quality of service (QoS) in deployment, of a resilient dense wavelength division multiplexing (DWDM) backbone, for global networks requires critical design-phase planning optimisation. A genetic algorithm (GA) model has been developed to solve the routing and wavelength assignment problem using binary variable-length chromosome encoding, under two different schemes of bandwidth optimisation (BOS) and delay optimisation (DOS). The performance of the new GA-based resiliency model has been evaluated for four benchmark networks: PAN EUROPEAN, COST239, NSFNET and ARPA2. Simulation results show, a superior capability and efficiency for the model to solve this complex, multi-constraint and nondeterministic polynomial-hard problem for BOS and DOS. The nonlinear nature of this process reveals a significant sensitivity for optical layer network topology on the optimum-design QoS. The results also demonstrate, that the PAN EUROPEAN network shows, the highest flexibility for primary path design, NSFNET for the secondary path and ARPA2 comes with the lowest design flexibility for both primary and secondary paths.