A New Network Traffic Prediction Approach in Software Defined Networks

Abstract

Software Defined Networking (SDN) is a centralized management network architecture, the handling commands of flows are designed in the controller and installed into flow tables of OpenFlow switches. SDN has obtained a lot of attention due to flexible and scalable. Network traffic prediction is very important for load balancing and network planning. It is implemented to improve the quality of service of the operators. In this paper, we propose a network traffic prediction method based on Short Time Fourier Transform (STFT) and traffic modeling. We use STFT to decompose network traffic into high-frequency components and low-frequency components. The low-frequency component of network traffic describes the smoothness and long-range correlation of network traffic, we model it as Auto-regression (AR) model. Otherwise, the high-frequency component of the network traffic fluctuates strongly which shows the randomness of the network traffic, we model the network traffic as an exponential distribution. However, since the prediction error of network traffic model is large, we propose an optimization function to optimize the predictions of network traffic to reduce the errors. Finally, we conduct some simulations to verify the proposed measurement scheme. From simulations, our proposed prediction method outperforms WABR and PCA.