Placement of wavelength converters and light splitters in a WDM network using the generic graph model

Abstract

Previous studies on the placement issue mainly focus on the availability of wavelength converters or light splitters on each node and normally the placement of wavelength converters and light splitters are considered in isolation. In this paper, by extending our previous work on the generic graph model, we study the placement of wavelength converters and light splitters as well as how their capabilities will affect the cost efficiency and performance in a holistic manner. In our study, both the sparse and limited capability strategies are considered, so is the coexistence of both wavelength converters and light splitters in the same network. The dynamic multicast traffic model is used for the simulation study. Four proposed algorithms, namely, WCF-I, WCF-D, LSF-I and LSF-D, for finding the optimal configuration of wavelength converters and light splitters are studied. The configuration includes information on the availability as well as the capability (and limitation) of the devices on each node in the network. Three metrics: cost efficiency (CE), request blocking probability (RBP) and average multicast tree cost (AMTC) are used to compare the different algorithms. All of the proposed algorithms manage to find a placement configuration with higher CE and comparable RBP, compared to the case of wavelength converters and light splitters having full capability of wavelength conversion and light splitting on all the nodes. From the results, we conclude that the presence of wavelength converters contributes more to improving the cost efficiency than the presence of light splitters when traffic load is not high. We also conclude that the sparse strategy has more impact on the performance than the limited strategy. In addition, our results led us to believe that the configuration that is obtained for a specific traffic load is not necessary the optimal configuration for other traffic loads.