Distributed State Monitoring for IaaS Cloud with Continuous Observation Sequence

Cloud computing has become increasing popular by freeing users from the low-level task of setting up the hardware and managing the system software. Anomaly detection is an effective approach to enhancing availability and reliability of Cloud infrastructures. In this paper, we propose a supervised online anomaly detection scheme that analyses monitoring data within current sliding data window and judging the current working state of the monitored component in Cloud based on Hidden Markov Model (HMM). What makes our method different from existing anomaly detecting is that it determines current state in context of recent continuous monitoring data, instead of isolated data point. Besides, our algorithm is basically distributed and runs locally on each computing machine on the Cloud in order to achieve high scalability. Experiments performed on real data sets validate the fact that our algorithm can effectively detect performance anomalies while imposing low overhead to the infrastructure in Cloud.