PASER:加性多维KPI异常根因定位模型

PASER:Root Cause Location Model for Additive Multidimensional KPIs

利用多维属性关键性能指标(key performance indicators,KPI)的可加性特征,能够实现对大型互联网服务故障的根因定位.由一项或多项异常根因导致的KPI数据变化,会导致大量相关KPI数据值的变化.提出一种基于异常相似性评估和影响力因子的剪枝搜索异常定位模型(pruning search model based on anomaly similarity and effectiveness factor for root cause location,PASER),该模型以多维KPI异常传播模型为基础,提出了衡量候选集合成为根因可能性的异常潜在分数评估方案;基于影响力的逐层剪枝搜索算法,将异常根因的定位时间降低到了平均约5.3 s.此外,针对异常根因定位中所使用的时间序列预测算法的准确性和时效性也进行了对比实验,PASER模型在所使用的数据集上的定位表现达到了0.99的F-score.

Additivity of multidimensional KPIs(key performance indicators)was used to achieve root cause location for large-scale Internet services.The anomaly caused by one or more root causes usually results in the change of a large number of relevant KPIs.A pruning search model based on anomaly similarity and effectiveness factor for root cause location(PASER)was proposed,which indicated the probability of candidate set becoming root cause using potential score based on the anomaly propagation model of multidimensional KPI.The pruning search algorithm used in PASER also managed to reduce the location time to about 5.3 seconds on average.In addition,the selection of time series prediction algorithm was also discussed.PASER had finally achieved a performance of 0.99 F-score on the experimental dataset.