Data-driven Anomaly Detection Method for Monitoring Runtime Performance of Cloud Computing Platforms

Cloud computing platforms are complex system, which consist of a lot of software working together. Because of software defects, cloud computing platforms may has performance anomaly during runtime. In this paper, a data-driven anomaly detection method is proposed to monitor runtime performance for cloud computing platforms. The proposed method can not only detect the performance anomaly of cloud computing platforms during runtime, but also find out which performance metric results in the anomaly. A series of experiments are conducted on a real private cloud computing platform based on OpenStack and experimental results show the proposed method is better than previous anomaly detection methods for cloud computing platforms.