Skip to Main Content
Linear optimization has been exhaustively used for resource conservation and network design/planning. The primary assumption facilitating the use of linear optimization is that traffic is assumed to be deterministic, which in reality is not the case. In the domain of optical networks, the routing and wavelength assignment (RWA) problem for provisioning lightpaths or optical circuits is well known and is solved using constrained linear optimization. However, with emerging service requirements such as triple play, video-on-demand (VoD) and pseudo-wire edge-to-edge emulation (PWE3), the optimization involves stochastic parameters for dynamic service provisioning. Alternate solutions to lightpath communication that solve the paradox between maximizing dynamism as well as maintaining high-efficiency are proposed. In the electronic domain, solutions such as resilient packet rings (RPR) and next generation packet-over-SONET (NG-POS) have been considered. Conversely, in the all-optical domain light-trails, have been proposed as a solution to meet requirements of these emerging services. Optimal allocation of resources and subsequent network design is not possible using traditional linear programming approaches due to the time-variant nature of traffic and architectural properties of these optical and electronic solutions. Stochastic optimization -a technique that assumes probabilistic nature of traffic is a promising approach for planning and allocation of resources in a network. This paper discusses application of stochastic optimization to high-speed all-optical networks, in particular, to the emerging concept of light-trail networks. Light-trails are a generalized lightpath that enable dynamic sharing of a wavelength leading to efficient network utilization. A stochastic formulation for design of light-trail networks in metropolitan rings is presented. A simpler (tractable) formulation is also developed using a quantization hypothesis that restricts the solution space. T- he formulation is then abstracted to the well known Bender's decomposition method for solving stochastic optimization problems. The formulation is evaluated under varying traffic conditions. Results are obtained and compared with linear optimization. Cost savings in terms of network equipment (transponders) is presented.