Efficient energy-aware controller placement in software-defined wireless sensor networks

Abstract

A centralized controller in the Software-defined Wireless Sensor Networks (SDWSN) environment poses a single point of failure and is inapt for a large-scale network. As leverage, multiple controllers have been introduced but are confronted with controller placement problems (CPP) for a better quality of service and network requirements. CPP challenge in SDWSN lies in finding the numbers, location and allocation of controllers in given network topology as well as sensors assignments. This is important in positively impacting the network's performance in terms of latency and cost minimization and reliability, and energy efficiency maximization. Moreover, several Software-defined networking (SDN) based CPP approaches have been proposed and developed over the years but only a few proposed techniques addressed energy efficiency in the SDWSN. Therefore, an efficient and dynamic CPP approach that is generic and considers energy consumption is important in the SDWSN. In this research, a hybrid central CPP algorithm is designed and developed to reduce or get rid of the wireless sensor network (WSN) and SDN-based network performance objectives for improved SDWSN network performance. The proposed algorithm considered energy consumption, propagation latency and cost metrics to prolong the lifetime of wireless sensors and minimize the delay and cost spent for placements of controllers in networks. The algorithm is associated with real controllers and wireless sensor devices that use certain types of modules. Furthermore, the technique utilized the threshold-sensitive energy efficient sensor network (TEEN) routing protocol and particle swarm optimization (PSO)- K-means algorithm chosen after empirical evaluations were performed with other protocols and algorithms. The approach is evaluated through a series of simulations and the results indicate that the proposed efficient CPP energy-aware algorithm is effective on SDWSN in terms of number, location and allocation of controllers compared to the traditional SDN and WSN. The proposed algorithm also outperformed other algorithms and significantly increases propagation latency. The proposed algorithm minimizes delay and improves energy consumption however, it is short on reliability and load balancing which is part of the future work. We, therefore, recommend using real or emulated CC2530 devices to create an open-source architecture and framework that can be used on network simulation tools to test centrally designed algorithms in SDWSN.