This paper studies the problem of light splitter placement (LSP) and wavelength converter placement (WCP) in all-optical WDM networks to enable optimal provisioning of static and dynamic traffic through efficient photonic multicast connections. To solve the LSP-WCP problem under static traffic provisioning, an Integer Linear Programming model is formulated to achieve the optimal solution in the sense that the total number of wavelength channels required by the multicast requests is minimized. To solve the LSP-WCP problem under dynamic traffic provisioning, a complementary-combined LSP-WCP heuristic is proposed to minimize the multicast traffic blocking probability, and is proved through extensive simulations.

Using less wavelengths to serve more communication channels is one of the primary goals in the design of WDM networks. By installing wavelength converters at some nodes in a network, the number of wavelengths needed can be reduced. It has been observed that the more converters installed in a network, the less number of wavelengths is needed, given the same network load. In this paper, we study the relationship between the number of converters and the number of wavelengths needed in a system, and propose a suite of theories and results on how to place the minimal number of converters in the system so that the number of wavelengths W is at most a constant α times the maximal link load L (i.e., W α L), where α = 3/2 or 5/3. The results show a significant saving of converters in networks of both special topologies and general topology.

In optical networks, wavelength converters are required to improve the efficiency of wavelength-division multiplexing. In this paper, we propose a genetic algorithm to determine the optimal locations of the nodes in the network where a given number of converters are placed. Optimality is achieved by the minimum wavelength blocking probability. Our algorithm is applied to two realistic networks constructed from the locations of major cities in Ibaraki Prefecture and from those in Kanto District in Japan and is shown to reach the nearly optimal solution in a limited number of generations. The accuracy is verified by simulation. The computational time is compared with that of an exhaustive search algorithm.