All Products
Search

This Product

    Elastic Compute Service:Configure an auto provisioning group

    Document Center

    Elastic Compute Service:Configure an auto provisioning group

    Last Updated:Apr 27, 2026

    Configure an auto provisioning group to create ECS instances across resource pools for scenarios such as machine learning and web services.

    Configure in the ECS console

    Example 1: Implement machine learning

    You plan to complete a week-long machine learning task that analyzes risk factors in mortgage loans. The instance cluster must meet the following requirements:

    • Region: China (Hangzhou).

    • Instances use NVIDIA V100 GPUs with up to 32 GB of GPU memory per instance.

    • Target capacity: 20 instances.

    • To minimize costs, create only spot instances. The cluster size can be less than the target capacity.

    • Release instances after the task completes.

    The following table lists the auto provisioning group settings for this scenario.

    Section

    Parameter

    Description

    Capacity Configuration

    Target Capacity

    Set the target capacity:

    • Select Instances from the Target Capacity drop-down list.

    • Enter 20.

    Instance Configuration

    Instance Configuration

    Select instance types with NVIDIA V100 GPUs and up to 32 GB of GPU memory:

    1. Select the ecs.gn6v-c8g1.2xlarge and ecs.gn6e-c12g1.3xlarge instance types.

    2. Check resource availability for the ecs.gn6v-c8g1.2xlarge and ecs.gn6e-c12g1.3xlarge instance types in Hangzhou Zone H and Hangzhou Zone I.

      Note

      You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.

    Add instance configurations by zone and instance type:

    1. Add an instance configuration:

      • Specify a vSwitch in Hangzhou Zone H.

      • Add the ecs.gn6v-c8g1.2xlarge and ecs.gn6e-c12g1.3xlarge instance types.

    2. Add another instance configuration:

      • Specify a vSwitch in Hangzhou Zone I.

      • Add the ecs.gn6v-c8g1.2xlarge and ecs.gn6e-c12g1.3xlarge instance types.

    These configurations produce the following resource pools:

    • The ecs.gn6v-c8g1.2xlarge instance type in Hangzhou Zone H

    • The ecs.gn6e-c12g1.3xlarge instance type in Hangzhou Zone H

    • The ecs.gn6v-c8g1.2xlarge instance type in Hangzhou Zone I

    • The ecs.gn6e-c12g1.3xlarge instance type in Hangzhou Zone I

    Provisioning Policy

    Select Cost Optimization Policy. The group creates instances from the lowest-priced resource pool.

    Spot Instance Interruption Settings

    Select Release. The cluster size can fall below the target capacity to minimize costs.

    Advanced

    Group Type

    Select One-time Delivery. The cluster size can fall below the target capacity to minimize costs.

    Start Time and End Time

    Set the start time and expiration time to cover the next week.

    Global Maximum Price for Spot Instances

    Select Set Maximum Price and specify your maximum acceptable price.

    Instance Shutdown Settings

    • Select Shut Down Instances Upon Group Expiration to release instances after the task completes.

    • Select Shut Down Excessive Instances When Target Capacity Is Exceeded to minimize costs.

    Example 2: Build a ticketing website

    You want to build a ticketing website that provides reliable service at all hours, especially during peak hours. The instance cluster must meet the following requirements:

    • Region: China (Hangzhou).

    • Each instance has no more than 8 vCPUs.

    • Target capacity: 80 vCPUs.

    • Minimum capacity: 60 vCPUs.

    • Optimize access experience at minimum cost based on the minimum capacity.

    • The cluster must support disaster recovery.

    The following table lists the auto provisioning group settings for this scenario.

    Section

    Parameter

    Description

    Capacity Configuration

    Target Capacity

    Set the target and minimum capacity:

    • Select vCPUs from the Target Capacity drop-down list.

    • Enter 80.

    • Select Use Pay-as-you-go Instances to Provide Computing Power.

    Pay-as-you-go Instance Capacity

    Enter 60 to meet the minimum capacity requirement.

    Instance Configuration

    Instance Configuration

    The c6 instance family suits frontend web servers. Select instance types with no more than 8 vCPUs:

    1. Select the ecs.c6.large, ecs.c6.xlarge, and ecs.c6.2xlarge instance types.

    2. Check resource availability for the ecs.c6.large, ecs.c6.xlarge, and ecs.c6.2xlarge instance types in Hangzhou Zone H, Hangzhou Zone I, and Hangzhou Zone J.

      Note

      You can go to the ECS Instance Types Available for Each Region page to view the instance types available in each region.

    Add instance configurations by zone and instance type:

    1. Add an instance configuration:

      • Specify a vSwitch in Hangzhou Zone H.

      • Add the ecs.c6.large, ecs.c6.xlarge, and ecs.c6.2xlarge instance types.

    2. Add another instance configuration:

      • Specify a vSwitch in Hangzhou Zone I.

      • Add the ecs.c6.large, ecs.c6.xlarge, and ecs.c6.2xlarge instance types.

    3. Add another instance configuration:

      • Specify a vSwitch in Hangzhou Zone J.

      • Add the ecs.c6.large, ecs.c6.xlarge, and ecs.c6.2xlarge instance types.

    These configurations produce the following resource pools:

    • The ecs.c6.large instance type in Hangzhou Zone H

    • The ecs.c6.xlarge instance type in Hangzhou Zone H

    • The ecs.c6.2xlarge instance type in Hangzhou Zone H

    • The ecs.c6.large instance type in Hangzhou Zone I

    • The ecs.c6.xlarge instance type in Hangzhou Zone I

    • The ecs.c6.2xlarge instance type in Hangzhou Zone I

    • The ecs.c6.large instance type in Hangzhou Zone J

    • The ecs.c6.xlarge instance type in Hangzhou Zone J

    • The ecs.c6.2xlarge instance type in Hangzhou Zone J

    Provisioning Policy

    Select Balanced Distribution Policy. The group evenly distributes instances across zones to improve disaster recovery.

    Instance Configuration

    Spot Instance Interruption Settings

    Select Stop to ensure low-latency website access. Recovering a stopped instance in economical mode is faster than creating a new instance.

    Advanced

    Group Type

    Select Continuous Delivery and Maintain Capacity for uninterrupted ticketing service.

    Start Time and End Time

    The group starts immediately and runs indefinitely for continuous ticketing service.

    Global Maximum Price for Spot Instances

    Select Automatic Bidding to maintain a stable instance count for optimal access experience.

    Instance Shutdown Settings

    Select Shut Down Excessive Instances When Target Capacity Is Exceeded to minimize costs.

    The target capacity uses vCPUs, so each instance type's weight equals its vCPU count. The following table shows the weighted price of each instance type.

    Note

    Prices in the following table are for reference only. Actual prices on the buy page prevail.

    Instance type

    vCPUs

    Pay-as-you-go price (USD)

    Weight

    Weighted price (USD)

    ecs.c6.large

    2

    0.06 per hour

    2

    0.03 per hour

    ecs.c6.xlarge

    4

    0.121 per hour

    4

    0.03025 per hour

    ecs.c6.2xlarge

    8

    0.241 per hour

    8

    0.030125 per hour

    The group first applies the balanced distribution policy to evenly distribute instances across zones, then selects instance types with lower weighted prices. If weighted prices are equal, instance types are chosen at random.

    Configure by calling an API operation

    Example 1: Use multiple resource pools with the lowest prices

    To minimize costs and reduce the impact of spot instance reclamation from a single resource pool, configure the group to use multiple lowest-priced resource pools.

    Call the CreateAutoProvisioningGroup operation with SpotAllocationStrategy set to lowest-price and SpotInstancePoolsToUseCount specified. The following example creates a group with these results:

    • Obtain five resource pools based on the lt-bp1ivgo4p5now3px**** launch template.

    • Set the target capacity to 30 instances, all spot instances.

    • Use the three lowest-priced resource pools, creating 10 spot instances from each.

    Parameter

    Example

    Description

    TotalTargetCapacity

    30

    Sets the target capacity to 30 instances.

    SpotTargetCapacity

    30

    Creates 30 spot instances.

    PayAsYouGoTargetCapacity

    0

    Creates no pay-as-you-go instances.

    SpotAllocationStrategy

    lowest-price

    Uses the cost optimization policy for spot instances.

    PayAsYouGoAllocationStrategy

    lowest-price

    Uses the cost optimization policy for pay-as-you-go instances.

    SpotInstancePoolsToUseCount

    3

    Uses the three lowest-priced resource pools.

    LaunchTemplateId

    lt-bp1ivgo4p5now3px****

    The launch template ID.

    LaunchTemplateConfig.1.VSwitchId

    vsw-bp1ygryo03m39xhsy****

    The vSwitch ID of extended configuration 1 (Resource Pool 1).

    LaunchTemplateConfig.1.InstanceType

    ecs.c6e.large

    The instance type of extended configuration 1 (Resource Pool 1).

    LaunchTemplateConfig.1.WeightedCapacity

    1

    The weight of extended configuration 1 (Resource Pool 1).

    LaunchTemplateConfig.2.VSwitchId

    vsw-bp16hgf8f3kvtcbyu****

    The vSwitch ID of extended configuration 2 (Resource Pool 2).

    LaunchTemplateConfig.2.InstanceType

    ecs.c6e.xlarge

    The instance type of extended configuration 2 (Resource Pool 2).

    LaunchTemplateConfig.2.WeightedCapacity

    1

    The weight of extended configuration 2 (Resource Pool 2).

    LaunchTemplateConfig.3.VSwitchId

    vsw-bp1oeawdo9tj2gvjp****

    The vSwitch ID of extended configuration 3 (Resource Pool 3).

    LaunchTemplateConfig.3.InstanceType

    ecs.c6e.2xlarge

    The instance type of extended configuration 3 (Resource Pool 3).

    LaunchTemplateConfig.3.WeightedCapacity

    1

    The weight of extended configuration 3 (Resource Pool 3).

    LaunchTemplateConfig.4.VSwitchId

    vsw-bp1oeawdo9tj2gvjp****

    The vSwitch ID of extended configuration 4 (Resource Pool 4).

    LaunchTemplateConfig.4.InstanceType

    ecs.g6e.xlarge

    The instance type of extended configuration 4 (Resource Pool 4).

    LaunchTemplateConfig.4.WeightedCapacity

    1

    The weight of extended configuration 4 (Resource Pool 4).

    LaunchTemplateConfig.5.VSwitchId

    vsw-bp1oeawdo9tj2gvjp****

    The vSwitch ID of extended configuration 5 (Resource Pool 5).

    LaunchTemplateConfig.5.InstanceType

    ecs.g6e.2xlarge

    The instance type of extended configuration 5 (Resource Pool 5).

    LaunchTemplateConfig.5.WeightedCapacity

    1

    The weight of extended configuration 5 (Resource Pool 5).

    Example 2: Preferentially use a specific resource pool

    Configure the group to prioritize a specific resource pool. For example, if you use reserved instances applicable to ecs.c6e.large instances, prioritize ecs.c6e.large to benefit from reserved instance discounts.

    Call the CreateAutoProvisioningGroup operation with PayAsYouGoAllocationStrategy set to prioritized and the target resource pool's priority set to 0 (highest). The following example creates a group with these results:

    • Obtain five resource pools based on the lt-bp1ivgo4p5now3px**** launch template.

    • Set the target capacity to 20 instances: 10 spot and 10 pay-as-you-go.

    • Use the ecs.c6e.large resource pool (LaunchTemplateConfig.1) for pay-as-you-go instances.

    Parameter

    Example

    Description

    TotalTargetCapacity

    20

    Sets the target capacity to 20 instances.

    SpotTargetCapacity

    10

    Creates 10 spot instances.

    PayAsYouGoTargetCapacity

    10

    Creates 10 pay-as-you-go instances.

    SpotAllocationStrategy

    lowest-price

    Uses the cost optimization policy for spot instances.

    PayAsYouGoAllocationStrategy

    prioritized

    Uses the priority-based policy for pay-as-you-go instances.

    LaunchTemplateId

    lt-bp1ivgo4p5now3px****

    The launch template ID.

    LaunchTemplateConfig.1.VSwitchId

    vsw-bp1ygryo03m39xhsy****

    The vSwitch ID of extended configuration 1 (Resource Pool 1).

    LaunchTemplateConfig.1.InstanceType

    ecs.c6e.large

    The instance type of extended configuration 1 (Resource Pool 1).

    LaunchTemplateConfig.1.WeightedCapacity

    1

    The weight of extended configuration 1 (Resource Pool 1).

    LaunchTemplateConfig.1.Priority

    0

    The priority of extended configuration 1. 0 indicates the highest priority.

    LaunchTemplateConfig.2.VSwitchId

    vsw-bp16hgf8f3kvtcbyu****

    The vSwitch ID of extended configuration 2 (Resource Pool 2).

    LaunchTemplateConfig.2.InstanceType

    ecs.c6e.xlarge

    The instance type of extended configuration 2 (Resource Pool 2).

    LaunchTemplateConfig.2.WeightedCapacity

    1

    The weight of extended configuration 2 (Resource Pool 2).

    LaunchTemplateConfig.3.VSwitchId

    vsw-bp1oeawdo9tj2gvjp****

    The vSwitch ID of extended configuration 3 (Resource Pool 3).

    LaunchTemplateConfig.3.InstanceType

    ecs.c6e.2xlarge

    The instance type of extended configuration 3 (Resource Pool 3).

    LaunchTemplateConfig.3.WeightedCapacity

    1

    The weight of extended configuration 3 (Resource Pool 3).

    LaunchTemplateConfig.4.VSwitchId

    vsw-bp1oeawdo9tj2gvjp****

    The vSwitch ID of extended configuration 4 (Resource Pool 4).

    LaunchTemplateConfig.4.InstanceType

    ecs.g6e.xlarge

    The instance type of extended configuration 4 (Resource Pool 4).

    LaunchTemplateConfig.4.WeightedCapacity

    1

    The weight of extended configuration 4 (Resource Pool 4).

    LaunchTemplateConfig.5.VSwitchId

    vsw-bp1oeawdo9tj2gvjp****

    The vSwitch ID of extended configuration 5 (Resource Pool 5).

    LaunchTemplateConfig.5.InstanceType

    ecs.g6e.2xlarge

    The instance type of extended configuration 5 (Resource Pool 5).

    LaunchTemplateConfig.5.WeightedCapacity

    1

    The weight of extended configuration 5 (Resource Pool 5).