Introducing Amazon CloudWatch Anomaly Detection

Finding, isolating, and troubleshooting issues with resources and applications is often a reactive exercise where operators respond to fired alarms by digging into multiple metrics and dashboards to find the problematic resources. Without prior domain knowledge, users can find it difficult to differentiate between normal versus problematic behavior. Applications that exhibit rapid growth, cyclical, or seasonal behavior, such as requests that peak during the day and taper off at night, are difficult to monitor using static thresholds. Amazon CloudWatch Anomaly Detection is a new feature that applies machine-learning algorithms to continuously analyze system and application metrics, determine a normal baseline, and surface anomalies without requiring user intervention. Customers can now identify runtime issues sooner, reducing system and application downtime. CloudWatch Anomaly Detection also adapts to metric trends, enabling customers to monitor the dynamic nature of system and application behavior with alarms that auto-adjust to situations such as time-of-day utilization peaks. This helps identify unexpected changes that result from connectivity issues, code change deployments, database errors, and other operational issues so they can be identified early and remediated quickly.