Abstract:

The virtual network control plan usually set the virtual link bandwidth to the maximum value of traffic to meet user's dynamically changing bandwidth requirements, which causes a waste of resources to a certain extent. In response to the problem, a virtual network topology reconfiguration approach based on hybrid traffic prediction is proposed. The hybrid traffic prediction algorithm based on parameter optimization selection is used to predict the network traffic in the next period, and the topology reconfiguration is performed according to the traffic prediction result, which avoids the ping-pong effect and saves more bandwidth resources. In order to improve the accuracy and efficiency of the traffic prediction algorithm, the wavelet decomposition method is first used to decompose the traffic data into high-frequency detailed time series and low-frequency approximate time series, and then the phase-space reconstruction method based on the particle swarm optimization is used to perform feature extraction on the time series to construct training samples. After that, the chaotic model is used to train and predict the detailed time series, and the extreme learning machine (ELM) neural network is used to train and predict the approximate time series. The simulation results show that the proposed traffic prediction algorithm guarantees the prediction accuracy, the running time is shorter, and the prediction efficiency is higher, thus ensuring that the topology reconfiguration approach can save more bandwidth resources.

Key words: traffic prediction, topology reconfiguration, wavelet decomposition, phase-space reconstruction, extreme learning machine (ELM)