Hybrid Optimization With Deep Spiking Equilibrium Neural Network for Software Defined Network Based Congestion Prevention Routing

Abstract

In the Internet of Things (IoT) network, within a period of time it constructs a huge network of millions and billions of things that incorporate with each other, leading to various technical and application problems. Also, on-time delivery of packets is significant in software-defined networks. In the current software-defined network, the bandwidth overhead is not considered when an enormous amount of traffic enters the network; this may lead to network congestion. To overcome this issue, a novel method is proposed to detect network congestion based on hybrid optimization, namely, the Hybrid Gannet with Pelican Optimization (HGPO) algorithm. The node-level congestions and the link-level congestions, which means the buffer overflow and the multiple nodes trying to utilize the channel at the same time, must be controlled efficiently. Once the network congestion gets controlled, the shortest route path selection and congestion prevention are performed perfectly with the aid of the proposed Deep Spiking Equilibrium Neural Network (DSENN). A few route discovery frequency vectors, such as the interroute discovery time and route discovery time of each node, are determined to prevent congestion. Finally, it is implemented in the Python platform successfully, and the achieved throughput, delay, packet loss ratio, packet delivery ratio, end-to-end delay, and performance measures of the proposed method are 140 Mbit/s, 17 ms, 3.8%, 0.35 s, and 100%, respectively, which outperforms the other compared traditional algorithms.