Strong formulations and cutting planes for designing digital data service networks

This paper deals with the problem of designing a least-cost digital data service (DDS) network that connects a given set of locations through digital switching offices with bridging capabilities. We present several alternative mixed 0-1 integer programming formulations and evaluate analytically their relative strengths by comparing their respective linear programming relaxations. By exploiting the structures inherent in a particularly strong formulation, we develop several classes of valid inequalities and cutting planes in order to tighten the initial formulation. For several problems of real-world data, computational results show that the strong formulation with valid inequalities and cutting planes generates a very tight lower bound (over 98% of the optimality) and so finds an optimal solution well within an acceptable time bound.