Anomaly Detection for Nodes Under the Cloud Computing Environment

Abstract

Due to the services diversity and dynamic deployment, the anomalies will occur on nodes under cloud computing environment. If a single node generates an anomaly, the associated nodes are affected by the abnormal node, which will result in anomaly propagation and nodes failure. In this paper, a method of anomaly detection for nodes under the cloud computing environment is proposed. Firstly, the node monitoring model is established by the agents deployed on each node. Secondly, the comprehensive score is used to identify abnormal data. The anomaly of the single node is judged by the time window-based method. Then, the status of directly associated nodes is detected through normalized mutual information and the status of indirectly associated nodes is detected through the node attributes in the case of a single node anomaly. Finally, other associated nodes affected by the abnormal node are detected. The experimental results showed that the method in this paper can detect the anomalies of single node and associated node under the cloud computing environment effectively.