All Products
Document Center

Auto Scaling:Overview

Last Updated:Nov 23, 2023

You can use the vertical scaling feature to upgrade or downgrade the instance types of pay-as-you-go instances. During peak hours, you can upgrade existing instance types to ensure sufficient computing power. During off-peak hours, you can downgrade existing instance types to reduce resource costs. Auto Scaling supports scheduled vertical scaling tasks and CloudMonitor metric-based vertical scaling tasks to monitor business workloads. By executing vertical scaling tasks, Auto Scaling can flexibly meet your business requirements.


  • Peak hours are predictable. You can create scheduled vertical scaling tasks. For example, your company runs a large number of tasks at the beginning of each month or runs a promotion in a specified period.

    In this case, you can create a scheduled vertical scaling task. This ensures sufficient computing power during the peak hours and minimizes resource costs during the off-peak hours.

  • Your business workloads change constantly. You can create CloudMonitor metric-based vertical scaling tasks. For example, you run a news business. You cannot predict break news.

    In this case, you can create a CloudMonitor metric-based vertical scaling task. Auto Scaling monitors the specified CloudMonitor metric. When the metric value falls outside the allowed threshold range, Auto Scaling executes the vertical scaling task to adjust instance types based on business requirements.


    After a vertical scaling task is executed, you must restart the instances to put the new instance types into effect. During the restart period, the service availability may be compromised.


  • Efficiency

    The vertical scaling feature enables automatic upgrade or downgrade of existing Elastic Compute Service (ECS) instance types in a scaling group. This saves you from creating or releasing ECS instances and making custom configurations.

  • Convenience

    You do not need to manually perform instance type adjustments, including stopping the instance, upgrading or downgrading the instance types, and restarting the instance. The vertical scaling feature does these tasks on your behalf.

  • Flexibility

    The vertical scaling feature supports scheduled adjustment of instance types and metric-based adjustment of instance types. You can determine which type of vertical scaling tasks to create based on your business requirements.

  • Cost-effectiveness

    The vertical scaling feature allows you to downgrade your instance types during off-peak hours to reduce resource costs. The vertical scaling feature also reduces the O&M costs resulting from continuous monitoring of resource usage.

Billing methods

The vertical scaling feature is free of charge, but the billing of instances changes after instance types are changed. Prices increase after instance types are upgraded and decrease after instance types are downgraded. For more information about the billing rules for pay-as-you-go instances, see Pay-as-you-go.


  • The vertical scaling feature changes only instance types of pay-as-you-go instances.

  • You can specify an instance type that is of the same instance family as the desired instance type in a vertical scaling task.

    For example, you can change the instance type of an instance from ecs.g6.large to ecs.g6.2xlarge, but not to instance types of other instance families such as ecs.c6.2xlarge and ecs.r6.2xlarge.

  • You can specify the type of vertical scaling task as upgrade or downgrade, not both. If you have different business requirements, you can configure multiple vertical scaling tasks.

  • If a vertical scaling rule is repeatedly executed during multiple cycles or based on a threshold, the following rules apply:

    • If you want to upgrade instance types, you can specify up to 10 instance types in ascending order of computing power (vCPUs and memory size).

    • If you want to downgrade instance types, you can specify up to five instance types in descending order of computing power (vCPUs and memory size).

  • When you configure a vertical scaling task as a recurring task, the interval between consecutive cycles must be greater than 30 minutes. Otherwise, frequent changes of instance types affect service availability.