The CloudWatch service collects the metrics and the logs from any systems it monitors. Each and every AWS service that we create and use in our account automatically starts reporting its metrics to CloudWatch. If the default metrics are not sufficient to paint a complete image of the state of our application, we have the ability to create custom metrics and send any kind of relevant custom data to the service.

Any metrics, logs, or custom data that CloudWatch receives are stored into the CloudWatch repository, which is stored within the region that the services operate in. This means that any data stored in one region will never be replicated to any other region. Any on-premises or remote services where we enable log delivery to CloudWatch will also have to be configured with one of the CloudWatch regional endpoints which will, of course, mean that the data will be stored within that region.

To retrieve the metrics, logs, and statistics from the CloudWatch environment, we can use the CloudWatch management console, the AWS CLI, and the SDKs. We will of course have to select a region in any of those tools due to the previously mentioned regionally bound metric storage.

We also have the ability to perform actions upon certain conditions called alarms. A CloudWatch alarm can send a notification or inform other services that a certain metric has breached a threshold, and the recipient can then perform some kind of manual or automated action to remediate the situation.

The following diagram represents the CloudWatch architecture: