Abstract

In this thesis we study the problem of designing a cost-efficient multicommodity flow network with survivability features. We study the geometrical structure of several polyhedra arising in this context. For some of these polyhedra we are able to give a complete description by extreme points and by facets, while for others we are able to give a complete description by extreme points and we present several classes of facet-defining inequalities. Initial testing with real data using a cutting-plane algorithm has shown that these inequalities are extremely effective in reducing the gap between the LP relaxation value and the optimal solution value.