Paper detail

Modeling the guaranteed delivery of bulk data

The delivery of bulk data is an increasingly pressing problem in modern networks. While in some cases these transfers happen in background without specific constraints in terms of delivery times, there are a number of scenarios in which the transfer of tens of GB of data must occur in specific, limited time windows. In order to face this task, a suitable solution is the deployment of virtual links with guaranteed bandwidth between endpoints provided by a Service Overlay Network (SON) provider. We model this scenario as an optimization problem, in which the target consists of minimizing the costs of the virtual links provided by the SON and the unknowns are the provisioned bandwidths of these links. Since the resulting objective function is neither continuous nor convex, the solution of this problem is really challenging for standard optimization tools in terms of both convergence time and solution optimality. We propose a solution based on an heuristic approach which uses the min-plus algebra. Numerical results show that the proposed heuristic outperforms the considered optimization tools, whilst maintaining an affordable computation time.