云化虚拟化网络业务可用度多参数敏感性分析  

Multi-parameter sensitivity analysis of network function virtualization application availability

作　　者：朱杰 黄宁[1,2] 程亮 ZHU Jie;HUANG Ning;CHENG Liang(School of Reliability and Systems Engineering,Beihang University,Beijing 100191,China;Yunnan Innovation Institute of Beihang University,Kunming 650233,China;Data Communication Products Line of Huawei Technologies Corporation Limited,Beijing 100191,China)

机构地区：[1]北京航空航天大学可靠性与系统工程学院,北京100191 [2]北京航空航天大学云南创新研究院,云南昆明650233 [3]华为技术有限公司数据通信产品线,北京100191

出　　处：《系统工程与电子技术》2022年第8期2677-2687,共11页Systems Engineering and Electronics

基　　金：国家自然科学基金(61872018)资助课题。

摘　　要：网络功能的虚拟化使得网络中软硬件设备类型变多,不同的设备有不同的故障修复参数,这些参数均会对业务可用度造成影响。由于不同的业务共享网络资源难以找到对业务可用度影响最大的关键设备参数,现有的局部敏感性分析方法难以保证结果的正确性。全局敏感性分析方法需要大量的仿真样本,操作困难,所以本文将网络中的参数分为异类和同类,对异类参数采用全效应指数法,对同类参数采用动态业务介数法,最终综合两种方法分析的结果找到网络中的关键参数。该方法能够保证结果准确性的同时极大减少仿真的样本量。Hardware and software device types are increasing with the virtualization of network functions. Different devices have different fault and recovery parameters, which affect application availability. It is difficult to find the key parameters that have the greatest impact on application availability, because different applications share network device resources. The correctness of the results are difficult to be ensured by existing local sensitivity analysis methods, and existing global sensitivity analysis methods require a large number of simulation samples and are difficult to be conducted. Therefore, this paper divides the parameters in the network into heterogeneous and homogeneous ones, and then analyses them by the total effect index method and the dynamic application intermediate method respectively. Finally, the key parameters in the network are found by synthesizing the analysis results of the two methods. The proposed method can ensure the accuracy of the results and greatly reduce the sample size of simulation.

关 键 词：云化虚拟化网络 业务可用度 敏感性分析 关键参数 

分 类 号：N94-0[自然科学总论—系统科学]