| 摘要: |
| 针对当前虚拟网络功能（Virtualization Network Functions，VNF）需求预测方法准确率较低且不适用于边缘网络的问题，提出了一种在边缘网络中基于支持向量回归（Support Vector Regression，SVR）与门控循环单元（Gated Recurrent Unit，GRU）神经网络模型结合的VNF需求预测方法。考虑到网络边缘流量具有突发性、自相似性及长相关性等特点，结合SVR和GRU两种模型的优点，利用计算复杂度较低的SVR和GRU模型分别提取网络服务历史时序数据的短期特征和长期特征，以提高VNF需求预测准确率，实现边缘网络中VNF的提前放置。实验表明，所提出的预测方法在边缘网络中针对不同网络服务的预测较于传统方法、循环神经网络（Recurrent Neural Networks，RNN）、长短期记忆网络（Long Short-Term Memory，LSTM）模型能够降低20%~30%的误差，有更佳的预测效果。 |
| 关键词: 边缘网络 虚拟网络功能 支持向量回归 门控循环单元 |
| DOI： |
|
| 基金项目:国家重点研发计划项目（2019YFB2102001） |
|
| A VNF demand forecasting method in edge networks |
| HUANG Hongcheng,BAO Xiaomeng,HU Min |
| (School of Communications and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China) |
| Abstract: |
| For the problem that the current virtual network function(VNF) demand forecasting methods have low accuracy and are not suitable for the edge network,this paper proposes a VNF demand forecasting method based on the combination of support vector regression(SVR) and gated recurrent unit(GRU) neural network model in the edge network,and analyzes the placement of VNF when the edge resources are sufficient or insufficient.According to the characteristics of network edge traffic,such as burst,self-similarity and long-term correlation,through combining the advantages of SVR and GRU models,SVR and GRU models with lower computational complexity are used to extract the short-term and long-term characteristics of network service historical time series data,so as to improve the accuracy of VNF demand prediction and realize the advance placement of VNF in edge network.Experiments verify that the proposed prediction method can reduce the error by 20% ~ 30% compared with the long-term memory network(LSTM) for different network services in the edge network,and has better prediction effect. |
| Key words: edge network virtual network function support vector regression gated recurrent unit |
