语义匹配在OpenStack故障诊断中的应用

摘要

在已有的OpenStack故障模式库基础上,提出基于日志异常特征的语义匹配故障诊断方法.首先,使用动态时间窗口方法,聚集属于同一故障所抛出的异常信息.其次,计算不同异常属性与异常类型的互信息,并采用语义匹配方法计算不同属性的近似值.最后,通过综合计算得到匹配度较高的故障模式,并构建OpenStack故障语义匹配系统原型.通过实验分析可知,基于日志异常特征的语义匹配故障诊断方法对于提升故障模式识别能力具有显著作用.%On the foundation of existing OpenStack fault library,a fault diagnosis method of semantic matching based on log anomaly feature is proposed.Firstly,using a dynamic time window,the anomaly information thrown by the same faults is aggregated.Secondly,the mutual information of different anomaly and exception types is calculated,and the approximation of different attributes with semantic matching is also calculated.Lastly,the fault pattern with high matching degree is obtained by comprehensive calculation and the semantic-based fault diagnosis system prototype of OpenStack is constructed.The experimental results show that the semantic-matching fault diagnosis method based on log anomaly feature has a significant effect on the improvement of fault pattern recognition.