Monitoring AWS Resources with Amazon CloudWatch

I recently completed the AWS SimuLearn: Resource Monitoring lab and learned how Amazon CloudWatch can monitor AWS resources, visualize metrics, detect problems, and trigger automated actions.

The main workflow I learned was:

A metric is a value measured over time, such as CPU usage, network traffic, or memory usage.

CloudWatch organizes metrics into namespaces:

A dimension identifies what a metric belongs to.

My mental model is:

Amazon EC2 automatically publishes several infrastructure metrics to CloudWatch.

For operating-system-level metrics such as memory usage, the CloudWatch Agent can collect data from inside the EC2 instance.

An application or CloudWatch Agent can also publish custom metrics through the CloudWatch API.

For an EC2 instance to push metrics, I learned that it generally needs:

For example, the IAM role may allow:

Using an IAM role is better than storing long-term access keys directly on the EC2 instance.

CloudWatch Dashboards combine multiple graphs and monitoring widgets into one place.

A widget can also keep its own time range, such as Last 3 hours, instead of following the dashboard's global time range.

A CloudWatch Alarm evaluates a metric against a condition.

For example:

The alarm can then trigger an action:

This means engineers do not need to watch the dashboard continuously. CloudWatch can detect the problem and respond automatically.

The simplest mental model I took away from this lab is:

CloudWatch is more than a graphing tool. It provides a centralized way to collect, visualize, detect, alert, and respond to changes in AWS resources and applications.