Photon Netw Commun (2009) 18:287–299 DOI 10.1007/s11107-009-0192-z A bicriteria routing model for multi-fibre WDM networks T. Gomes · J. Craveirinha · J. Clímaco · C. Simões Received: 14 November 2008 / Accepted: 3 February 2009 / Published online: 25 February 2009 © Springer Science+Business Media, LLC 2009 Abstract All-optical WDM networks are characterised by multiple metrics (hop-count, cost, delay, available band- width, loss probability, reliability), but generally routing algorithms only optimise one metric. Having in mind the inherent limitations of this type of approach, it seems poten- tially advantageous in the development of multicriteria models capable of explicitly representing the different per- formance objectives and enabling to treat in a consistent man- ner the trade off among the various criteria. A bicriteria model for obtaining a topological path (unidirectional or symmetric bidirectional) for each lightpath request in a WDM network is proposed. The first criterion is related to bandwidth usage in the links (or arcs) of the network. The second criterion is the number of links (hops) of the path. The model resolution approach uses a k -shortest path algorithm as well as prefer- ence thresholds defined in the objective function space, com- bined with a Chebyshev distance to a reference point (which changes with the analysed preference region). The solution of this bicriteria model is a non-dominated topological path. A heuristic procedure is then used to assign wavelengths to the links. The performance of the bicriteria model is analysed by comparing it with two monocriterion approaches, using two test networks. Keywords Multicriteria · Routing in WDM networks T. Gomes (B ) · J. Craveirinha · J. Clímaco University of Coimbra, INESC Coimbra, Coimbra, Portugal e-mail: teresa@deec.uc.pt C. Simões Polytechnic Institute of Viseu, INESC Coimbra, Coimbra, Portugal 1 Introduction and motivation In a wavelength-routed WDM network, node pairs can establish point-to-point all-optical connections, or lightpaths, for information transfer. In the absence of wavelength con- verters, a lightpath must use the same wavelength on all the links of its route (the wavelength continuity constraint), but wavelengths can be reused by different lightpaths in the net- work, as long as they do not share any fibre link. Given a set of connection requests, the problem of setting up lightpaths by defining a path and assigning a wavelength to each of its links for every connection is called the Routing and Wavelength Assignment (RWA) problem. When the networks nodes are equipped with wavelength converters, a lightpath can use different wavelengths on dif- ferent links along the route. Obviously, wavelength conver- sion leads to lower blocking probabilities, but, in practice, some studies have shown that with only a small number of converters placed in strategic locations, a significant per- formance improvement can be achieved [16, 51]. A wave- length converter is said to have a conversion degree D, if it can shift any wavelength to one of D wavelengths. Another approach for improving the average number of established lightpaths is to use several fibres per link. In the absence of wavelength converters, multi-fibre networks also have to satisfy the wavelength continuity constraint. However, the chances of finding the same wavelength free on all the links along the path is higher, as it can choose the free wavelength on any of the fibres in a link. A multi-fibre network with F fibres per link and W wavelengths per fibre is functionally equivalent to a single-fibre network with F × W wavelengths and conversion degree of F [3]. Typically, the representation of connection requests may be of three types: static, incremental and dynamic [53]. In the case of static traffic, the entire set of connections is known in 123