Sun Wenwen, Guan Shaopeng
School of Information and Electronic Engineering, Shandong Technology and Business University, Yantai, Shandong, China.
School of Electronic and Information Engineering, Huaibei Institute of Technology, Huaibei, Anhui, China.
PeerJ Comput Sci. 2022 Jun 23;8:e1011. doi: 10.7717/peerj-cs.1011. eCollection 2022.
With the continuous development and improvement of Software-Defined Networking (SDN), large-scale networks are divided into multiple domains. Each domain, which is managed by a controller, forms multi-domain SDN architecture. In multi-domain SDN, the dynamics and complexity are more significant, bringing great challenges to network management. Comprehensively and accurately predicting traffic situation in multi-domain SDN can better maintain network stability. In this article, we propose a traffic situation prediction method based on the gated recurrent unit (GRU) network in multi-domain SDN. We first analyzed the relevant factors that affect data traffic and control traffic and transformed them into a time series of actual situation values. Then, to enhance the prediction performance of GRU, we used the salp swarm algorithm to optimize the hyperparameters of GRU automatically. Finally, we adopted hyperparameter optimized GRU to achieve traffic situation prediction in multi-domain SDN. The experimental results indicate that the proposed method outperforms other traditional machine learning algorithms in terms of prediction accuracy.
随着软件定义网络（SDN）的不断发展和完善，大规模网络被划分为多个域。每个由控制器管理的域构成了多域SDN架构。在多域SDN中，动态性和复杂性更为显著，给网络管理带来了巨大挑战。全面准确地预测多域SDN中的流量状况能够更好地维护网络稳定性。在本文中，我们提出了一种基于门控循环单元（GRU）网络的多域SDN流量状况预测方法。我们首先分析了影响数据流量和控制流量的相关因素，并将它们转化为实际状况值的时间序列。然后，为了提高GRU的预测性能，我们使用鹈鹕群算法自动优化GRU的超参数。最后，我们采用超参数优化后的GRU实现多域SDN中的流量状况预测。实验结果表明，所提方法在预测准确性方面优于其他传统机器学习算法。
