A Survey on Anomaly Detection Techniques in Large-Scale KPI Data

For Internet-based services quality, it is very necessary for Internet companies to monitor a large number of key performance indicators (KPIs) and accurately detect anomalies. With the increasingly complex structure of the system, the changing characteristics of the performance monitoring data have gradually become a challenge for anomaly detection. Recently, in the performance management sector, there has been renewed interest in research on anomaly detection of KPI streams. There has been a lot of work in the area of clustering-based unsupervised anomaly detection. This paper presents a survey of various clustering-based anomaly detection techniques and discusses the advantages, limitations, and practical significance of different algorithms. Some practical application-related kinds of literature are summarized. At the end of the paper, we put forward some new research trends and opinions and suggestions for the research direction.