Bandwidth allocation with differential pricing for flexible demands in data center networks

This article addresses the problem of bandwidth sharing in data center networks. A promising approach is the use of advance bandwidth reservation systems. However, reservation systems are generally based on deterministic models that assume users to have precise knowledge of their demands, which is unlikely. Deviating from this, we propose a new model that allows a user to specify flexible bandwidth demands. A provider (in this model) is to reserve minimum bandwidth for the entire duration of an accepted request, while allocating additional bandwidth for a fraction of the duration, such that the fraction is within a ‘flexibility range’ specified in the request. We tie up the model with differential pricing, and formulate bandwidth allocation as a two-phase optimization problem. The solution to the problem defines bandwidth profiles for accepted requests while maximizing revenue for providers. We show that problems in both phases are NPNP-hard, and develop computationally fast algorithms for the two phases. Numerical studies demonstrate that, in comparison to the deterministic model, our model brings down the number of rejected requests significantly, while increasing revenue for providers.