Performance Improvement of IP Networks under Internet of Things Traffic using Deep Learning and Genetic Algorithm

The quality of delivering Internet of Things (IoT) traffic in IP networks is of great importance in IoT era. In this article, a Genetic Algorithm (GA)-based method is proposed to select the routing and scheduling strategy of each IP router to improve the quality of service of IoT traffic. To this aim, we first propose a method based on deep learning to distinguish IoT traffic from none-IoT ones. The trained model results in 99% accuracy on test data. Then, distinct scheduling and routing methods are suggested for these two traffic types in network routers. The aim of GA-based strategy selection is to improve the latency and reliability of IoT traffic without compromising the performance of none-IoT ones. Here, we utilize a set of scheduling algorithms including FIFO, Fair, Weighted Fair, and Priority algorithms to construct GA chromosomes. Also, a set of routing algorithms, i.e., Dijkstra, A*, BFS, and DFS are used in definition of chromosomes. Simulation results demonstrate that the proposed method leads to a significant improvement in latency and reliability of IoT traffic.