On Solving the Rooted Delay- and Delay-Variation-Constrained Steiner Tree Problem

We present mixed integer programming approaches for optimally solving a combinatorial optimization problem arising in network design with additional quality of service constraints. The rooted delayand delay-variation-constrained Steiner tree problem asks for a cost-minimal Steiner tree satisfying delay-constraints from source to terminals and a maximal variation-bound between particular terminal path-delays. Our MIP models are based on multi-commodity-flows and a layered graph transformation. For the latter model we propose some new sets of valid inequalities and an efficient separation method. Presented experimental results indicate that our layered graph approaches clearly outperform the flow-based model.