Capacity Planning of Survivable Mesh-based Transport Networks under Demand Uncertainty

摘要

Almost all existing work on the design of survivable networks is based on a specific demand forecast to which one optimizes routing and transport capacity assignment for a single target planning view. In practice these single-forecast models may be used repetitively by a planner to consider a range of different scenarios individually, hoping to develop intuition about how to proceed. But this is not the same as having a planning method that can inherently and quantitatively consider a range of possible futures all at once. Our approach considers both the cost of initial design construction and the expected cost of possible augmentations or "recourse" actions required in the future, adapting the network to accommodate different actual future demands. In practice, these recourse actions might include lighting up a new DWDM channel on an existing fiber or pulling-in additional cables, or leasing additional capacity from third party network operators, and so on. A stochastic linear programming approach is used to achieve designs for which the total cost of current outlays plus the expected future recourse costs is minimized. Realistic aspects of optical networking such as network survivability based on shared spare capacity and the modularity and economy-of-scale effects are considered. These are not only important practical details to reflect in planning, but they give the "future-proof design problem for such networks some unique aspects. For instance, what is the working capacity under one future scenario that may not waste capacity if that demand scenario does not materialize, because the same channels may be used as shared spare capacity under other future scenarios. Similarly economy-of-scale effects bear uniquely on the future-proof planning problem, as the least-cost strategy on a life-cycle basis may actually be to place more capacity today than current requirements would suggest. This is of obvious relevance to planners given the recent hard times in the telecommunications industry, causing a tendency to minimize costs now regardless of the consequences.