Metaheuristic-Based Approaches for Solving the Controller Placement Problem in Software-Defined Wireless Sensor Networks (SDWSNs)

Software-Defined Networking (SDN) is well-known as a master solver for many traditional networks restrictions. Network management flexibility is the SDN core characteristic that recently becomes a hot topic for many researchers to improve network performance. A critical issue arises specifically for the multi-controller SDN-based network, namely the controller placement problem (CPP), known as an NP-hard problem. Solving such problems in a reasonable amount of time is usually carried out using metaheuristic algorithms. This thesis presents different approaches for solving the CPP which are based on metaheuristic algorithms such as the genetic algorithm (GA) and the greedy randomized adaptive search procedure (GRASP). The first approach is a fault tolerance metaheuristic-based scheme (FTMBS) that we proposed for solving the CPP in wireless software-defined networks. The FTMBS is a multi-objective-based scheme aiming to maximize network connectivity and the load balance among controllers, minimize the network worst-case latency, and maximize the network lifetime in the presence of faulty nodes. In the presence of conflicting multi-objective metrics, the decision-maker or the network administrator decides on the compromise between these conflicting metrics. We defined the selection criteria for the number of SDN controllers ahead of time that achieves the targeted average percentage of network improvement. Simulations carried out for three and five controllers showed that three controllers were enough for such networks. We have shown the efficiency of the proposed FTMBS scheme under various percentages of faulty nodes as it has lower latency compared to the random distribution of controllers among cluster heads and the cluster-based network partition algorithm (CNPA). Also, we have verified the goodness of the solutions by showing that the GA solutions have a good approximation to the Pareto optimal solutions provided by the Non-dominating Sorting Genetic Algorithm (NSGA-II). In addition, the proposed FTMBS scheme showed its superiority over various state-of the-art schemes for different network performance metrics on the cost of having a bit higher complexity. The second approach, namely the ALBATROSS scheme, is an energy-efficient strategy for WSNs, which is a modification of the FTMBS scheme. The albatross bird's dynamic soaring scheme is adopted in the cluster heads selection algorithm, and the selected cluster heads are taken as inputs for the heuristic algorithm to solve the CPP. Simulation results showed that the ALBATROSS scheme saves the network energy and outperforms other existing energy-aware schemes found in literature. However, it has a bit higher complexity than other schemes. Besides, in Appendix A, we have provided an example of using the SDN technology in a practical environment other than WSNs, such as wireless body area networks (WBANs). Keywords: Software-defined networking, controller placement problem, meta heuristic algorithm, network performance enhancement.