An Anomaly Detector Deployment Awareness Detection Framework Based on Multi-Dimensional Resources Balancing in Cloud Platform

Anomaly detection has been an important topic in cloud platforms to guarantee the dependability and robustness of services in the cloud. Most research works were proposed to tackle the detection performance problems of detection algorithms caused by the volume of data, the dynamic environment, various types of anomalies, and so on. However, almost all of them take only the optimization of algorithms into account, which leads to a situation that some key features of detector deployment and the scalability and dependability of the detection framework itself are omitted. Therefore, an anomaly detector deployment awareness detection framework based on multi-dimensional resources balance is proposed to address the problems. It balances the multi-dimensional resources by bringing four factors of resources into consideration to deploy detectors quietly, which are the current utilizations, the available capacity, the demands of detectors and the remaining resources. Three experiments and comparative analysis suggest that this framework achieves a higher scalability and detection accuracy than the existing framework.