Posted On: Nov 19, 2021
Amazon CloudWatch now supports anomaly detection based on metric math expressions. Amazon CloudWatch anomaly detection allows you to apply machine-learning algorithms to continuously analyze system and application metrics, determine a normal baseline, and surface anomalies with minimal user intervention. CloudWatch metric math allows you to aggregate and transform metrics to create custom visualizations of your health and performance metrics. Metric math supports basic arithmetic functions such as +,-,/,*, comparison and logical operators such as AND & OR, and a number of additional functions such as RATE and INSIGHT_RULE_METRIC. For example, with AWS Lambda metrics you can divide the Errors metric by the Invocations metric to get an error rate, use anomaly detection to visualize expected values on a metric graph, and create an anomaly detection alarm to dynamically alert you when the value falls outside of the expected range.
It is easy to get started with anomaly detection for metric math. In the CloudWatch console, go to Alarms in the navigation pane to create an alarm based on anomaly detection, or start with metrics to overlay the math expression’s expected values onto the graph as a band. You can also enable anomaly detection using the AWS Command Line Interface, AWS SDKs, or AWS CloudFormation templates.