Deploy a model

Billing

This feature consumes DataWorks resources and other cloud service resources. The fees are described below.

Fee type	Description
DataWorks fees	A running model service in DataWorks consumes resources from a serverless resource group. For pay-as-you-go resource groups, you are charged based on CU-hours. For more information, see Billing of serverless resource groups.
Non-DataWorks fees	PrivateLink instance fees and request processing fees. For more information, see Billing of PrivateLink. Private Hosted Zone fees. For more information, see Billing of Private Hosted Zone.

Usage notes

DataWorks Model Service runs in a fully managed mode within DataWorks resource groups. For secure private network access, you must enable the relevant services during deployment. The system automatically configures PrivateLink and Private Hosted Zone for the following reasons:

PrivateLink enables cross-VPC access, allowing you to reach the fully managed model service from a VPC within your account that can communicate with the DataWorks resource group.
Private Hosted Zone provides custom domain name resolution within your account, allowing you to access a DataWorks Model Service by domain name within your VPC.

When you invoke a model service by domain name, traffic flows from your VPC through the PrivateLink channel to the DataWorks resource group's VPC, reaching the model instances to complete the invocation. You can view service details in the PrivateLink console and the Alibaba Cloud DNS-Private Zone console.

Prerequisites

Enable the PrivateLink service.

The model service is fully managed on the DataWorks platform side. To access it from your VPC, use PrivateLink to establish a secure cross-VPC connection.
Enable the Private Hosted Zone service.

Private Hosted Zone is used to configure custom domain name resolution on your account. It forwards request traffic to the DataWorks Model Service when you access the model service by using a fixed domain name.
Create a DataWorks workspace and associate a serverless resource group with it.

Important

Make sure that the region where you activate the services is the same as the region of your DataWorks workspace. Otherwise, the services may not function properly.

Limitations

The Model Service feature is available only in the following regions: China (Hangzhou), China (Shanghai), China (Beijing), China (Ulanqab), China (Shenzhen), China (Hong Kong), Japan (Tokyo), Singapore, Malaysia (Kuala Lumpur), Indonesia (Jakarta), Germany (Frankfurt), US (Silicon Valley), and US (Virginia).
Only model deployment is supported. Model training is not supported.
Each tenant (Alibaba Cloud account) can deploy up to 50 model services per region.
This feature supports only serverless resource groups, and each resource group can have up to 5 model services deployed.
Each model service can be associated with up to 3 VPCs.

Go to the feature

Log on to the DataWorks console. In the top navigation bar of the console, switch to the desired region.
In the left-side navigation pane, click Model Service to go to the Model Service page.

On the Model Service page, click Deploy Model to go to the Model List page. Select the model that you want to deploy and click Deploy to go to the Model Deployment configuration page. Complete the model deployment based on the following instructions.

Configure basic information

Model: Confirm the type of model to be deployed to the DataWorks resource group. For more information, see Supported models.
Service Name: A custom name to identify this model service in the resource group.
Remarks: Optional. Add remarks to help identify and manage the model service.

Configure resource information

Configure the deployment environment for the model.

Parameter	Description
resource group	Select the serverless resource group where the model service will be deployed. After deployment, you can log on to the DataWorks console, switch to the desired region, and click Resource Group in the left-side navigation pane to go to the resource group list page. Click the target Resource Group Name and view the model service-related serverless resource group usage on the resource group Details page.
vSwitch	Specify the vSwitch environment where the model service is deployed. Follow the on-screen instructions to select a vSwitch that is associated with the serverless resource group in an applicable zone.
Deployment specification	Select the resource specification for each model service instance.
Instances	Set the number of instances for the model service. Multiple instances improve availability.
Total CU usage	The total CU usage for the model service is calculated based on Deployment Specification × Instances. Make sure that the available CUs in the resource group for model services are sufficient to deploy the current model service. You can go to the resource group quota management page to adjust the CU limit allocated to model services. For more information, see Manage resource group quotas. You can scale up a subscription resource group. The upper limit of a pay-as-you-go resource group is `500CU`. After a model is deployed, the system automatically adjusts the maximum upper limit of the pay-as-you-go resource group to `2000CU`.

After the configuration is complete, click Deployment to start the deployment.

Important

The system automatically creates PrivateLink, Alibaba Cloud DNS - Private Hosted Zone, and security group resources when creating a model service. Do not manually delete or modify these resources. When a model service is released, the system automatically cleans up the related resources.

After the deployment is complete, the system performs the following operations under your account:

Creates a PrivateLink endpoint in the default VPC associated with the DataWorks resource group and establishes an encrypted channel with the PrivateLink service in the DataWorks resource group VPC.
Automatically adds a DNS resolution record in the Alibaba Cloud DNS - Private Hosted Zone service under your account, associating the private domain name with the VPC environment of the DataWorks resource group.

Manage models

After a model service is created, you can manage its status, view service details, manage networks and API keys, and adjust resources in the Model Service list.

Manage model status

A model service starts automatically after creation. You can manage its status from the Operation column of the Model Service list.

Operation	Service status change	Resource consumption
Start	Running	Consumes resources from the serverless resource group.
Stop	Stopped	Does not consume resources from the serverless resource group.
Restart	Restarting → Running	The model service is briefly unavailable during the restart. After the restart is complete, it continues to consume resources from the serverless resource group.
Delete	——	Permanently releases resources from the serverless resource group.

View model service

The Overview tab displays the configuration of the current model service, including Basic Information, Resource Configurations, and Invocation Information.

In the Model Service list, find the target model service and click its Service Name to go to the Overview tab of the model service.
You can view the basic information, resource information, and invocation information of the model service.
- Basic information: Includes the model Service Name, Service ID, Model type, and other information.
- Resource information: Displays the Deployment Specification, Instances, and other information.
- Invocation information: To invoke a model service in a node task, click the icon next to VPC Endpoint Domain to copy the domain name parameter.

Modify model resources

You can modify the service name, adjust the deployment specification, and change the number of instances for a model service.

In the Model Service list, find the target model service and click its Service Name to go to the Overview tab of the model service.
Click Modify on the right side of Resource Configurations to go to the Modify Resources page and configure the settings.

Important
Resource changes will cause the service to restart and temporarily affect the model service.

Manage model networks

The Network Configuration tab lists the VPC environments that can access the model service over the private network. You can add or remove VPCs on this tab.

In the Model Service list, find the target model service and click its Service Name to go to the Overview tab of the model service.
Switch to the Network Configuration tab to view the VPC environments that can access the model service over the private network.
To extend the access scope, click Add Network to allow more VPC environments to access the model service deployed on DataWorks over the private network.

When adding a network, specify a VPC and a vSwitch. After the Status of the VPC changes to Enabled, you can access the model service through this VPC.
Note
- Billing: After you add a VPC to a model service, the system creates a PrivateLink endpoint in the VPC that you select to access the model service. This establishes a network connection with the DataWorks resource group. At the same time, Private Hosted Zone adds a private domain name resolution record. Instance fees, traffic processing fees, and domain name resolution fees are incurred during this process. For billing details, see Billing of PrivateLink and Billing of Private Hosted Zone.
- Limit: You can add up to 3 VPCs to a model service.
If you no longer want a VPC environment to access the model service, click Delete for the target VPC in the model service.

When you remove a VPC from the model service, the PrivateLink endpoint that was created in this VPC under your account is also removed.

Manage API keys

An API key is a credential used for identity and permission verification when invoking the model service. You can manage all API keys on the API KEY tab.

In the Model Service list, find the target model service and click its Service Name to go to the Overview tab of the model service.
Switch to the API KEY tab to create, manage, and use API keys:
- Add an API key: After deployment, the system creates a built-in API key for invocations from other DataWorks modules. To invoke the model service from other environments by using its Endpoint, click Add API Key to create a new API key.
  
  We recommend that you create separate API keys for different application scenarios.
- View an API key: Click View in the Operation column of the target API key, and then click Copy to obtain the API key.
- Delete an API key: DataWorks API keys support the Disable and Delete operations.
  
  Important
  If you want to Disable or Delete an enabled API key, evaluate the impact in advance. Once an API key is Disable or Delete, all tasks that use this API key to invoke the model will fail.
  
  The Disable or Delete operation takes approximately 5min to take effect.

View logs

On the model service details page, switch to the Logs tab to view runtime logs. Use these logs to troubleshoot invocation exceptions or performance issues.

Next step

After the model is deployed, you can invoke the model service to develop related tasks.

Invocation workflow

The following flowchart illustrates the invocation workflow of model services:

When you deploy a model service in a DataWorks resource group or configure a VPC for the model service, the system automatically performs the following operations:

Establishes cross-VPC connections. The system automatically creates a PrivateLink endpoint in your VPC (a VPC under your account that can communicate with the DataWorks resource group) and establishes an encrypted communication channel with the PrivateLink service in the DataWorks resource group VPC.

This operation automatically creates a PrivateLink endpoint under your account. Make sure that you have activated the PrivateLink service.
Configures domain name resolution. The system automatically configures domain name resolution rules in the VPC associated with the model service, so that domain name requests within the VPC are automatically forwarded to the DataWorks Model Service.

This operation automatically deploys Private Hosted Zone resolution under your account. Make sure that you have activated the Private Hosted Zone service.