Each metric has to be identified somehow, and with CloudWatch this is done through dimensions. Dimensions are key-value pairs that identify metrics. For example, each EC2 instance will have a dimension of instance-id that will have a value of that instance's ID.

Each metric can have up to 10 dimensions assigned to it. By adding additional dimensions we have the ability to present the data according to those additional dimensions. For example, when our application is composed of hundreds of instances that belong to different development environment such as production, test, and staging, it might be difficult to understand which parts of the application are consuming how many resources. We can project additional dimensions on the metrics from our environment to group and isolate different parts of an application in our data analysis, graphing, and reporting procedures.