Network Design for Multi-hour Traffic Profile

The level of traac and the distribution of traac tend to vary in a telecommunications network. Obtaining a minimum-cost network for multi-hour traac proole is a large-scale optimization problem. We formulate this as a linear programming problem and use this as a basis for comparing the eecacy of the heuristics we introduce. Our heuristic is to view this as a multi-period network planning problem where there is growth in aggregate traac from period to period, as well as redistribution of traac. Initial numerical experience indicates a difference of less than 10% between network cost obtained by this heuristic and that obtained by the large-scale linear programming formulation while the heuristic model is about N 2 t times faster to solve where N t is the number of time periods.