SOMBA-automated anomaly detection for Cloud quality of service

Cloud computing has transformed the standard model of service provisioning, allowing the delivery of on-demand services over the Internet. With its inherent requirements for elastic scalability and a pay-as-you-go pricing model, an additional level of complexity is added to its Quality of Service (QoS) management. This has made service provisioning more prone to performance anomalies due to the large-scale and evolving nature of Clouds. Existing methods for anomaly detection based on QoS monitoring in the Cloud rely on probabilistic methods, which are not computationally easy and are often valid for very short times before system dynamics change. We posit that more minimalistic approaches including automated techniques are needed for effective anomaly detection to support QoS enforcement in Clouds. In this paper, we present an automated anomaly detection scheme that recognises and adapts to changes in Clouds for efficient multi-metric performance anomaly detection to guarantee service quality. It includes a monitoring tool for collating performance data in real time for analysis and an anomaly detection technique based on an unsupervised machine learning strategy. Based on a Cloud service provisioning use case scenario, we evaluate our anomaly detection technique and compare it against two statistical anomaly detection approaches to demonstrate its efficiency.