Learning‐based multi‐constraint resilient controller placement and assignment insoftware‐defined networksusing covering graph

Abstract

AbstractThe controller placement problem is one of the most important issues in designing software‐defined networks (SDNs). In this context, one of the major challenges is the problem of multi‐constraint resilient controller placement and assignment (MCRCPA). In this article, considering the main criteria for solving the problem of MCRCPA, we propose a learning‐based approach named Lb‐MCRCPA, which includes a wide range of constraints including resiliency, switch‐to‐controller delay, switch‐to‐controller reliability, inter‐controller delay, and controller capacity in both main and backup modes. In the proposed method, we use an innovative structure called constraint covering graph (CCG) to calculate the reliability and to apply the set of constraints. The CCG is converted into a distributed learning model called irregular cellular learning automata (ICLA). The rules embedded in ICLA, together with the proposed assignment algorithm, can accurately determine the location of controllers. The performance evaluation of the proposed method for about 2600 different cases on more than 20 standard topologies of different sizes, not only proves its superiority over other methods but also proves its near‐optimality compared to the optimal solution.