面向多模态网络业务切片的虚拟网络功能资源容量智能预测方法  

Intelligent prediction method of virtual network function resource capacity for polymorphic network service slicing

作　　者：兰巨龙[1] 朱棣 李丹[1] LAN Julong;ZHU Di;LI Dan(Institute of Information Technology,Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学信息技术研究所,河南郑州450001

出　　处：《通信学报》2022年第6期143-155,共13页Journal on Communications

基　　金：国家重点研发计划基金资助项目(No.2020YFB1804803);国家自然科学基金资助项目(No.62002382)。

摘　　要：为解决多模态网络场景中网络切片上的VNF部署方案调整不及时造成的切片性能下降问题,提出了一种基于时空特征提取的VNF资源容量预测方法,旨在通过预测为即将到来的切片需求进行VNF的预部署。所提方法首先对用于预测的数据流时间序列进行加权处理,然后把处理后的时间序列及其依赖的空间拓扑信息输入网络模型中,通过图卷积网络重组时间序列的空间分布特征,再由门控循环单元感知输入数据的时序依赖关系,最后基于数据流序列与VNF实例数量的映射关系,由前馈神经网络最终输出VNF资源需求预测视图。实验结果表明,所提方法比对照方法中预测精度最高的方法提高了6.54%的需求预测精度。To solve the problem of network slicing performance degradation caused by untimely adjustment of VNF deployment scheme on network slicing in polymorphic network scenario,a VNF resource capacity prediction method based on spatial-temporal feature extraction was proposed,which aimed to pre-deploy VNF by forecasting for upcoming slicing needs.Firstly,the data flow time series used for prediction were weighted,and then the processed time series and its dependent spatial topology information were inputted into the network model.Then,the spatial distribution features of the time series were reorganized through the graph convolution network,and then the timing dependence of the input data was perceived by the gated recurrent unit.Finally,based on the mapping relationship between the data flow sequence and the number of VNF instances,the feed forward neural network outputted the VNF resource demand prediction view.The experimental results show that the proposed method improves the demand prediction accuracy by 6.54%over the comparison method with the highest prediction accuracy.

关 键 词：多模态网络 虚拟网络功能 资源容量 时空特征提取 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]