By default, all metrics created by any AWS service are collected on a 5-minute intervals. These are the so-called normal metrics. But AWS has enabled us to allow detailed metrics within a high number of services; the detailed metrics available from the AWS services are available in 1-minute intervals. The detailed metrics enable us to perform actions upon alarms from triggers that are much closer together than with the normal metrics. For example, when performing an action on two consecutive triggers, we will need to wait 10 minutes before the action is performed, while with detailed monitoring we can do it in only 2 minutes. While normal metrics are free, enabling detailed monitoring will mean that we will incur an additional cost for collecting the metrics with a higher frequency.

But is 1 minute detailed enough? Modern applications running in containers or as serverless Lambda functions can be scaled in under a second, and data that is minutes old might be too old. This is why CloudWatch enables us to collect custom metrics from our environment for a period as short as 1 second. Collecting metrics as frequently as 1 second apart will allow us to react to any kind of change in our application in real time.