Any application that consumes any kind of compute, storage, or network resources should be monitored. The term monitoring is commonly used in place of four separate features that allow us to get complete insight into the state of our applications:

• Resource usage metering
• System and application log collection
• Resource utilization graphing
• Alarming upon predefined trigger

We need to collect metrics and logs to be able to determine patterns of usage. The metrics, logs, and patterns will allow us to understand the resource consumption characteristics of an application, and once we have understood the way our application consumes resources, we can create alarms that are usually represented as certain thresholds, beyond which our application operates in an abnormal state.

Many monitoring systems are available as open source products or proprietary solutions. Some have the four previously mentioned characteristics, while others are very focused tools. Some monitoring systems also upgrade the four characteristics by providing us with mechanisms that can connect to resources and perform actions over these resources. While CloudWatch itself does not provide any actionable items, it can work together with other AWS services such as SNS, autoscaling, and Lambda to perform any kind of action that can be automated.

Amazon CloudWatch is a fully managed, fully integrated metering, log collection, graphing, and alarming utility for resources running in the AWS cloud and any on-premises resources connected to the AWS cloud.