Planning Self-Healing Ring Capacity under Demand Uncertainty

Abstract

We describe an approach to planning the capacity for self-healing rings in a survivable network when there is uncertainty about the demands that will be served. When detailed, reliable forecast data is not available, network plans based on stochastic methods that consider multiple demand scenarios are usually superior to those based on single point-estimates. However, a large number of potential realizations usually have to be considered before a reasonable plan can be developed. In addition, information regarding the distribution of the point-to-point demands may be elusive. In this paper, we show that when survivable components, like self-healing rings, are placed into the network some robustness to demand uncertainty is naturally provided, so the planning process can be simplified. We present an approach to sizing these components that uses the information available, like the total demand in a geographic area or the total demand terminating at each location. These data are often more reliable and easier to obtain than individual point-to-point demand values. We describe the differences between sizing unidirectional and bidirectional rings in the presence of demand uncertainty. Unlike unidirectional rings, where the capacity depends only on the total demand, the capacity requirement for bidirectional rings varies with the values of the individual demands and their locations. However, bidirectional rings are often more economical and more robust to demand uncertainty. In our approach, we employ a fast routing heuristic that find the capacity of bidirectional rings under a variety of demand scenarios. The scheme allows for a more accurate comparison between ring alternatives and the risks inherent to each. We describe how this approach to sizing can be incorporated into existing decision support for survivable network planning.