Datadog Releases Anomaly Detection To Facilitate Identification Of Abnormal Behavior In Cloud-Based Applications

Datadog today announces the release of Anomaly Detection, a machine learning-based tool that empowers engineering teams to expeditiously identify abnormalities within dynamic cloud environments. Anomaly Detection draws upon statistical analytics to pinpoint notable deviations in application performance behavior from its normal operating state. Designed to identify abnormal behavior within cloud-based applications that indigenously experience significant variations with respect to the scale of their underlying infrastructure as well as the velocity and variety of incoming data, Anomaly Detection creates a constantly updated representation of the normal state of an application that can identify aberrant behavior while taking into account a cloud-based application’s organic growth and volatility. As such, the application helps engineers and application owners understand genuinely abnormal application behavior by considering factors such as business growth, seasonality and macroscopic trends in the space in which the application operates that may compel varying levels of application behavior and performance. A key advantage of Anomaly Detection is its ability to absolve engineers of the task of updating metrics and thresholds in response to the evolution of an application over time. Datadog’s Anomaly Detection platform dovetails with its existing Outlier Detection tool that delivers alerts when a server behaves differently from analogues within its cohort. The combination of Anomaly Detection and Outlier Detection means that Datadog continues to strengthen its position as a leader in the cloud monitoring space by virtue of its impressive technology for automating the detection of aberrant application behavior. The graphic below illustrates Anomaly Detection’s ability to identify an abnormal deviation from innocuous variations in application behavior as shown by the red section of the line, which deviates significantly from the normal, sinusoidal trajectory of the graph.