Abstract

In this article, we analyze the effective management of user requirements in a software-defined network. The network uses customer segmentation slicing to provide different user needs. We consider two parameters for each segment: propagation of the latency between switches and controllers, and the latency between users and network access nodes. For this purpose, we propose a mixed-integer quadratic programming model subject to internal backbone and user connectivity constraints. The CPLEX solver is used to solve the quadratic and its linearized version models to obtain the optimal solutions. As far as we know, this is the first attempt to optimize user segmentation management in a software-defined network while taking into account latency as a critical quality of service requirement distinguishing different slices. In particular, we consider three slices of users so far. We assume that users of slice one can only connect to controller nodes. Whilst the remaining users can only connect to switch-Type nodes. We conduct preliminary numerical results on 13 real-world benchmark network instances. Our numerical experiments show that the linearized version of the quadratic model allows one to obtain the optimal solutions in significantly less CPU time which cannot be achieved with the quadratic model.