Traffic Prediction Based VNF Migration with Temporal Convolutional Network

In network function virtualization enabled networks with dynamic traffic, virtual network function (VNF) migration has been considered as an effective way to improve quality of service as well as resource utilization. However, due to time-varying network traffic, designing a fast and accurate VNF migration algorithm is still a great challenge. To address this issue, in this paper, we exploit the temporal convolutional network (TCN) to predict traffic flow for VNF migration decision in a fast and accurate manner. Based on the predicted results, we define a metric, i.e., migration index, to represent the load trend of each node in the network. A fast and efficient heuristic VNF migration algorithm is then proposed based on the migration index, with the goal to minimize the total migration cost in a time period. Extensive simulations are carried out to validate the effectiveness of TCN for traffic prediction. The results demonstrate that the proposed VNF migration algorithm can reduce the total migration cost up to 20% compared with existing algorithms.