Optimal regenerator assignment and resource allocation strategies for translucent optical networks

Abstract

Physical layer impairments in wavelength-routed networks limit the maximum distance, a signal can travel in the optical domain, without significant distortion. Therefore, signal regeneration is required at some intermediate nodes for long-haul lightpaths. In translucent WDM networks, sparsely located regenerators at certain nodes can be used to offset the impact of physical layer impairments. The routing and wavelength assignment (RWA) techniques in such translucent networks need to take into consideration the availability of regenerators and the maximum optical reach of the transparent lightpaths (without any regeneration). Although there has been significant research interest in RWA algorithms for translucent networks, much of the research has focused on dynamic RWA techniques. Only a handful of recent papers have considered the static (offline) case, and they typically propose heuristic algorithms to solve this complex design problem for practical networks. In this paper, we propose a generalized integer linear program (ILP) based formulation for static regenerator assignment and RWA in translucent WDM optical networks, with sparse regenerator placement. To the best of our knowledge, such a formulation that optimally allocates resources for a set of lightpaths for translucent networks, given the physical network, the locations of the regenerators, and the maximum optical reach has not been considered before. The proposed formulation is important for two reasons. First, it can serve as a benchbe developed in the future. Second, we show that using a novel node representation technique, it is possible to drastically reduce the number of integer variables. This means that unlike existing ILP formulations, our approach can actually be used to generate optimal solutions for practical networks, with hundreds of lightpath demands.

Impact of Wavelength and Modulation Conversion on Translucent Elastic Optical Networks Using MILP

“…In the first case, the nodes in the network that are equipped with regeneration circuits have been pre-selected. There has been quite some research recently on how to select regeneration nodes, including [12]. The constraints that the MILP must satisfy for all D sd ∈ D and Lij ∈ L are as follows:…”

“…When the allocation of regeneration nodes is unknown, we use the variable X ij,sd to differentiate the cases when node i is a regeneration node or not, as in Eq. (12). Constraints that limit the number of OEO circuits per regeneration node can also be included using:…”

Resource Assignment for Fiber Optic Networks

“…In the first case, the nodes in the network that are equipped with regeneration circuits has been pre-selected. There has been quite some research recently on how to select regeneration nodes, including [45]. The constraints that the MILP must satisfy for all D sd 2 D and L ij 2 L are as follows: The second case we consider is one where the regeneration nodes are not preselected.…”

“…Our formulation can implement modulation scheme selection, mid-lightpath modulation conversion (MC) and/or wavelength conversion (WC), and regeneration circuit allocation (to satisfy either a quality of service constraint or conversion function). MILP has previously been used to solve the resource assignment optimization problem in optical networks [10,41,42] and many other optical networking's key issues, such as: minimizing capital investment [1], [43], the regenerator placement problem [44], [45], network restoration [46], and network service capability maximization [47]. However, to the best of our knowledge, no published MILP solution has included these flexibilities in an optimal way for designing EONs.…”

“…As we mentioned in previous chapters, many works have applied the powerful mixed-integer linear programming (MILP) tool to solve optical networking issues: [1], [43], [44], [45], [46], [47]. However, in these and other previous work when researchers use LP to address network problems, they only consider a static tra c model.…”