Platform For AI:Elastic resource pool

Use the console

1. Log on to the PAI console. Select a region on the top of the page. Then, select the desired workspace and click Elastic Algorithm Service (EAS).

2. Click Deploy Service. In the Custom Model Deployment section, click Custom Deployment.

3. Navigate to the Resource Deployment section on the Custom Deployment page and configure the essential parameters. The table below outlines the critical configuration options. For comprehensive parameter details, refer to Model service deployment by using the PAI console.

   Parameter	Description
   Resource Type	Choose EAS Resource Group from the available options.
   Resource Group	Select an existing dedicated resource group from the dropdown list.
   Elastic Resource Pool	Enable the Elastic Resource Pool toggle and designate a public resource group as the Resource Type to activate elastic resource pooling for services deployed within your dedicated resource group. When Elastic Resource Pool is activated and your dedicated resource group reaches full capacity, the system automatically provisions pay-as-you-go instances in the public resource group during scale-out operations. These additional instances are billed according to public resource pricing, and during scale-in events, the system prioritizes releasing public resource instances first.

   

4. Complete the deployment by clicking Deploy.

Use the client

You can configure auto-scaling capabilities during service deployment using the EASCMD client. The following instructions demonstrate the process using a Windows 64-bit server as an example.

1. Configure a JSON file.

   Important

   Resource configuration methods and Virtual Private Cloud (VPC) direct connection capabilities differ depending on your service's resource group type. Within public resource groups, utilize the cloud.computing parameter to define specific node types and allocate additional resources for your service. The cloud.networking parameter enables VPC direct connectivity for your service. For services deployed in dedicated resource groups, VPC direct connections can only be configured at the resource group level. When deploying services in dedicated resource groups with elastic resource pooling enabled, you must configure the cloud.networking parameter to maintain VPC connectivity throughout service scaling operations.

   The following code provides sample content of the JSON file:

   {
     "model_path": "http://examplebucket.oss-cn-shanghai.aliyuncs.com/models/lr.pmml",
     "name": "test_burstable_service",
     "processor": "pmml",
     "metadata": {
       "instance": 1,
       "cpu": 1,
       "resource": "eas-r-xxx",
       "resource_burstable": true
     },
      "cloud": {
           "computing": {
               "instance_type": "ecs.r7.2xlarge"
           },
           "networking": {
               "security_group_id": "sg-uf68iou5an8j7sxd****",
               "vswitch_id": "vsw-uf6nji7pzztuoe9i7****"
           }
       }
   }

   In the preceding code:

   • resource_burstable: Controls auto-scaling activation for your service. Setting this parameter to true enables auto-scaling functionality.

   • cloud.networking: This parameter has no effect on services deployed within dedicated resource groups. When enabling elastic resource pooling for your service, configuring this parameter ensures continuous VPC direct connectivity throughout scaling operations.

   • cloud.computing: Optional parameter that allows you to define specific node types within the public resource group during scale-out operations. For detailed specifications, see Use public resources.

   For comprehensive parameter documentation, refer to JSON deployment.

2. Execute service deployment using the EASCMD client. Detailed instructions are available in Deploy a model by using the EASCMD client.

   When your dedicated resource group lacks sufficient capacity to accommodate service scale-out demands, newly added service instances will automatically utilize the public resource group.

Use the console

1. Log on to the PAI console. Select a region on the top of the page. Then, select the desired workspace and click Elastic Algorithm Service (EAS).

2. Select Update from the Actions column corresponding to your service.

3. In the Resource Information section of the Update Service page, enable or disable the resource auto scaling feature.

   • Enable resource auto scaling

     In the Resource Information section, activate the Elastic Resource Pool toggle and specify the public resource group type.

   • Disable resource auto scaling

     In the Resource Information section, deactivate the Elastic Resource Pool toggle.

4. Confirm your changes by clicking Update.

Use the client

# Enable the elastic resource pool feature for a deployed service. 
eascmdwin64.exe modify <service_name> -Dmetadata.resource_burstable=true

# Disable the elastic resource pool feature for a deployed service. 
eascmdwin64.exe modify <service_name> -Dmetadata.resource_burstable=false