面向数据中心网络的分层式故障诊断算法

A Hierarchical Fault Diagnosis Algorithm for Data Center Networks

首次提出一种面向数据中心网络的分层式故障诊断算法---HFD(Hierarchical Fault Diagnosis)算法.算法根据数据中心网络的结构特点,将网络系统分为两个逻辑层.在第一层,通过对"基本单元组"进行多次最简测试,将其划分为"相对故障单元组"与"相对正常单元组"两大类;在第二层,结合本文首次提出的"正-反关联性"思想和笔者曾经提出的"绝对故障基"思想,设计出AD-FAFD算法进行各单元间的相互测试与诊断.仿真实验结果表明,HFD算法在保持良好的诊断精确度的同时,有效降低了时间复杂度.

We first propose the hierarchical fault diagnosis algorithm-HFD algorithm for data center network.According to the structural characteristics of data center network,the algorithm divides the network system into two logical layers. On the first lay-er,we perform multiple simplest tests on the”basic unit groups”,which will be divided into”relative fault unit groups”and”rela-tive fault-free unit groups”.On the second layer,combining the“positive-negative correlation”idea firstly proposed in this paper and the“absolute fault base”idea previously proposed by the authors,we design the AD-FAFD algorithm for mutual test and diagnosis among units. The simulation result shows that HFD algorithm reduces the time complexity effectively while keeping good diagnosis accuracy.