Star capacity-aware latency-based next controller placement problem with considering single controller failure in software-defined wide-area networks

Abstract

The failure of a single controller in the control layer of the Software-Defined Network (SDN) challenges the management of switches in the data layer and disrupts the proper function of the network; therefore, the optimal placement of multiple controllers in the network and the assignment of switches to controllers, called Controller Placement Problem (CPP), is an emerging and very important issue. Because, in case of controller failures, switches must be reassigned to active controllers in the network, this process results in a significant increase in the worst-case latency after reassignment due to lack of planning for controller failures. For this reason, the problem can be customized for the network real world by planning the star assignment of switches to controllers despite failures. Also, for placing controllers, another challenge is the network search space, which can be significantly reduced by using standard array decision variables. In this paper, a new star capacity-aware delay-based controller placement approach with single controller failure is formulated and presented as a Mixed Integer Program (MIP). The purpose of the proposed approach is to minimize the maximum, for all switches, of the sum of the worst-case latency from the switch to the nearest first controller with enough capacity and the worst-case latency from the same switch to the closest second controller with enough capacity. We also use the population-based simulated annealing algorithm for fast convergence of the problem toward the optimal solution on large-scale networks. The proposed formulation and algorithm are estimated with the real Internet Topology Zoo and the results were extensively analyzed. Unlike controller placement approaches with linear (hierarchy) strategy, our proposed approach which is based on star planning and uses the array switch-to-controller assignment variables. According to the simulation results, our proposed approach performs better than Capacitated Next Controller Placement (CNCP) and Resilient Capacity-aware Controller Placement Problem (RCCPP) approaches in case of controller failure and in failure free case.