Arima for traffic load prediction in software defined networks

Abstract

Internet traffic prediction is needed to allocate and deallocate the resources dynamically and to provide the QoS (quality of service) to the end-user. Because of recent technological trends in networking SDN (Software Defined Network) is becoming a new standard. There is a huge change in network traffic loads of data centers, which may lead to under or over-utilization of network resources in data centers. We can allocate or deallocate the resources of the network by predicting future traffic with greater accuracy. In this paper, we applied two machine learning models, i.e., AR (autoregressive) and ARIMA (Autoregressive integrated moving average) to predict the SDN traffic. The SDN traffic is viewed as a time series. And we showed that the prediction accuracy of ARIMA is higher than the AR in terms of Mean Absolute Percentage Error (MAPE).