Application Auto Scaling concepts

An entity that you create to specify the resource that you want to scale. Each scalable target is uniquely identified by a service namespace, resource ID, and scalable dimension, which represents some capacity dimension of the underlying service. For example, an Amazon ECS service supports auto scaling of its task count, a DynamoDB table supports auto scaling of the read and write capacity of the table and its global secondary indexes, and an Aurora cluster supports scaling of its replica count.

When Application Auto Scaling automatically decreases capacity for a scalable target, the scalable target scales in. When scaling policies are set, they cannot scale in the scalable target lower than its minimum capacity.

When Application Auto Scaling automatically increases capacity for a scalable target, the scalable target scales out. When scaling policies are set, they cannot scale out the scalable target higher than its maximum capacity.

A scaling policy instructs Application Auto Scaling to track a specific CloudWatch metric. Then, it determines what scaling action to take when the metric is higher or lower than a certain threshold value. For example, you might want to scale out if the CPU usage across your cluster starts to rise, and scale in when it drops again.

The metrics that are used for auto scaling are published by the target service, but you can also publish your own metric to CloudWatch and then use it with a scaling policy.

A cooldown period between scaling activities lets the resource stabilize before another scaling activity starts. Application Auto Scaling continues to evaluate metrics during the cooldown period. When the cooldown period ends, the scaling policy initiates another scaling activity if needed. While a cooldown period is in effect, if a larger scale out is necessary based on the current metric value, the scaling policy scales out immediately.

Scheduled actions automatically scale resources at a specific date and time. They work by modifying the minimum and maximum capacity for a scalable target, and therefore can be used to scale in and out on a schedule by setting the minimum capacity high or the maximum capacity low. For example, you can use scheduled actions to scale an application that doesn't consume resources on weekends by decreasing capacity on Friday and increasing capacity on the following Monday.

You can also use scheduled actions to optimize the minimum and maximum values over time to adapt to situations where higher than normal traffic is expected, for example, marketing campaigns or seasonal fluctuations. Doing this can help you improve performance for times when you need to scale out higher for the increasing usage, and reduce costs at times when you use fewer resources.