Auto Scaling workflow

Updated at:
Copy as MD

Set up Auto Scaling by creating scaling groups, scaling configurations, and scaling rules, then trigger them manually or automatically.

Workflow

image
Note

If you select Launch Template or Select Existing Instance as the instance configuration source when creating a scaling group, Auto Scaling automatically creates and activates a scaling configuration. You can then enable the scaling group without creating one manually.

  1. Create a scaling group

    A scaling group is a set of identical instances for the same business scenario, with configurable minimum and maximum instance limits, scale-out templates, and scale-in policies. Overview of scaling groups.

  2. Create a scaling configuration

    A scaling configuration is a template that Auto Scaling uses to automatically launch ECS instances or elastic container instances. Overview.

  3. Enable the scaling configuration

    A scaling group can have multiple scaling configurations, but only one can be active at a time. If no scaling configuration is active after group creation, you are prompted to enable one. Manage scaling configurations.

  4. Enable the scaling group

    Scaling events run only in scaling groups in the Enabled state. When a scaling group has an active scaling configuration, you are prompted to enable the group. You can also enable it from the Scaling Groups page. Enable a scaling group.

  5. Create a scaling rule

    A scaling rule defines how to adjust the number of instances in a scaling group. Overview.

  6. Execute the scaling rule

    Execute a scaling rule using one of the following methods. Execute a scaling rule.

    • Manual execution: Run a scaling rule on demand for temporary business needs.

    • Automatic execution based on scheduled tasks: Execute a scaling rule at a specified time. Ideal for predictable workload patterns.

    • Automatic execution based on event-triggered tasks: Monitor metrics, report alerts, and execute a scaling rule when thresholds are breached. Ideal for unpredictable workload patterns.

Tutorial

Scenarios

Scenario description and example

References

You have temporary service requirements where the number of required servers and their deployment times are unpredictable. Because no monitoring metrics are available, you must manually adjust the number of servers.

For example, a company needs to add servers for testing to meet a temporary requirement. This requires you to manually control when and how many servers are added or removed.

Manually scale ECS instances with a few clicks

Your workload has predictable, cyclical fluctuations. You can anticipate peak and off-peak hours and want to automatically add or remove ECS instances at specific times.

For example, a gaming company experiences a consistent surge in its workload every night from 18:00 to 23:00, which requires additional ECS instances. During other periods, a smaller, fixed number of ECS instances is sufficient for daily operations.

Scheduled auto scaling for ECS instances

Service demand fluctuates, which makes it difficult to predict the required number of servers. You need to adjust server capacity in real time based on the workload.

For example, a news website experiences unpredictable traffic. When a popular news story breaks, traffic surges. After the story is no longer current, traffic drops.

Automatically scale ECS instances based on service workload