Identifying and diagnosing anomalies in application behavior is critical to delivering reliable
application-level performance. In this paper we introduce a strategy to detect anomalies and
diagnose the possible reasons behind them. Our approach extends the traditional window-
based strategy by using signal-processing techniques to filter out recurring, background
fluctuations in resource behavior. In addition, we have developed a diagnosis technique that
uses standard monitoring data to determine which related changes in behavior may cause …

Identifying and diagnosing anomalies in application behavior is critical to delivering reliable
application-level performance. In this paper we introduce a strategy to detect anomalies and
diagnose the possible reasons behind them. Our approach extends the traditional window-
based strategy by using signal-processing techniques to filter out recurring, background
fluctuations in resource behavior. In addition, we have developed a diagnosis technique that
uses standard monitoring data to determine which related changes in behavior may cause …