ECS instances can be provisioned by using multiple methods, such as individual provisioning, batch provisioning, high-availability deployment, and automatic cluster creation. These provisioning methods can be made in the ECS console or through API operation calls to meet your instance creation requirements in various scenarios.

Manually create instances

Scenarios: create multiple ECS instances of the same instance type and billing method in the same zone.

Creation methods:

Number of instances that can be created: If you create instances in the ECS console, the number of instances that can be created at a time depends on the usage amount of your ECS instances. If you create instances by calling the RunInstances operation, you can create 1 to 100 instances per call.

The following figure shows the instance lifecycle when you create ECS instances in the console or by calling the RunInstances operation:Lifecycle of instances created by the RunInstances operation

You can also call the CreateInstance operation to create an ECS instance. However, the instance enters the Stopped state after the instance is created, and you must manually start the instance.

Deploy instances on multiple physical servers to improve availability of the cluster (deployment sets)

Scenarios: deploy multiple ECS instances across physical servers. The instance creation mode is suitable for providing computing power to applications that require high availability and disaster recovery of underlying databases.

Creation method: create a deployment set first and then specify the deployment set when you create ECS instances. You can create instances in the ECS console or by calling the RunInstances or CreateInstance operation.

Number of instances that can be created: It depends on the creation method. If you create instances in the console or by calling the RunInstances operation, you can create one to seven instances at a time. If you create instances by calling the CreateInstance operation, you can create only one instance at a time.

Limits:
  • A maximum of seven ECS instances can be created for each deployment set in a zone.
  • Only specific ECS instance types are supported. For more information, see Overview.
  • Billing methods: Subscription and pay-as-you-go are supported. Preemptible instances are not supported.

Automatically create instance clusters at minimal costs (auto provisioning)

Scenarios: deploy instance clusters of different billing methods and instance types in different zones with one click. The instance creation mode is suitable for scenarios that require stable computing power to be provisioned quickly and preemptible instances to reduce costs.

Creation method: create an auto provisioning group, which automatically creates multiple ECS instances at a time.

Number of instances that can be created: 1 to 1,000 instances can be created for each auto provisioning group.

Limits: Pay-as-you-go instances and preemptible instances are supported. Subscription instances are not supported.

Procedure:

Automatically create and release instances (Auto Scaling)

Scenarios: maintain instance clusters of different billing methods and instance types in different zones. The creation mode is suitable for scenarios where service workloads fluctuate from time to time.

Creation method: create a scaling group and a trigger task. The scaling group automatically creates or releases multiple ECS instances at a time.

Number of instances that can be created:
  • A maximum of 1,000 ECS instances can be created within a scaling activity.
  • A maximum of 1,000 ECS instances can be created by using one scaling group.

Limits: Pay-as-you-go instances and preemptible instances are supported. You can manually add an existing subscription instance to a scaling group, but cannot create a subscription instance in the scaling group.

Auto Scaling also provides simplified functions which improves provisioning efficiency, shortening the time from requirement to provisioning. For example, you can configure an ECS instance to automatically associate itself with an SLB instance and an RDS instance. You can also configure lifecycle hooks to perform custom operations on the instance. You can use Auto Scaling to obtain the scalability that meets your business needs. For information about best practices of Auto Scaling, see: