Platform For AI:Elastic resource pool

Enable resource auto scaling in the console

1. Go to the Deploy Service page. For more information, see Model service deployment by using the PAI console.

2. In the Resource Deployment Information section of the Deploy Service page, configure the required parameters. The following table describes key parameters. For more information about other parameters, see Model service deployment by using the PAI console.

   Parameter	Description
   Resource Group Type	Select an existing dedicated resource group.
   Elastic Resource Pool	Turn on Elastic Resource Pool to enable the elastic resource pool feature for the services that are deployed in the dedicated resource group. Configure the instance type. For more information, see the Resource Configuration Mode section in the Model service deployment by using the PAI console topic. If you turn on Elastic Resource Pool and the dedicated resource group that you use to deploy services is fully occupied, the system automatically adds pay-as-you-go instances to the public resource group during scale-outs. The added instances are billed as public resources. The instances in the public resource group are preferentially released during scale-ins.

   

3. Click Deploy.

Enable resource auto scaling in the client

You can enable auto-scaling for a service when you deploy the service by using the EASCMD client. The following section describes the procedure by using a Windows 64 server as an example.

1. Configure a JSON file.

   Important

   The configuration methods of resources and the virtual private cloud (VPC) direct connection feature vary based on the resource group type of the service. In the public resource group, you can use the cloud.computing parameter to specify the required node type and obtain more resources for a service. You can also use the cloud.networking parameter to enable the VPC direct connection feature for the service. If a service is deployed in a dedicated resource group, you can enable the VPC direct connection feature only for the dedicated resource group. If you deploy a service in a dedicated resource group and enable the elastic resource pool feature for the service, you must configure the cloud.networking parameter to ensure the availability of VPC direct connections during service scaling.

   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: Specifies whether auto scaling is enabled for the service. If you set this parameter to true, auto scaling is enabled for the service.

   • cloud.networking: This parameter does not take effect for services that are deployed in dedicated resource groups. If you enable the elastic resource group feature for the service, you must configure this parameter to ensure the availability of VPC direct connections during service scaling.

   • cloud.computing: Optional. You can specify the required node type in the public resource group during a scale-out. For more information, see Work with the public resource group.

   For information about other parameters, see Parameters of model services.

2. Deploy the service by using the EASCMD client. For more information, see Deploy model services by using EASCMD or DSW.

   If a dedicated resource group is insufficient to support a service scale-out, the service instances added to the service use the public resource group.

Use the console

1. Go to the Elastic Algorithm Service (EAS) page. For more information, see Model service deployment by using the PAI console.

2. Click Update Service in the Actions column of the service.

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

   • Enable resource auto scaling

     In the Resource Deployment Information section, turn on Elastic Resource Pool and configure the instance type of the public resource group.

   • Disable resource auto scaling

     In the Resource Deployment Information section, turn off Elastic Resource Pool.

4. Click Deploy.

Use the EASCMD 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