Dedicated resource groups are classified into subscription dedicated resource groups and pay-as-you-go dedicated resource groups. This topic describes how to manage the two types of dedicated resource groups, including how to create, scale out, temporarily scale out, and renew dedicated resource groups, and how to enable the virtual private cloud (VPC) direct connection channel. The lifecycles and billing methods of the two types of resource groups are different. You must select the proper type based on your needs.

Background information

Compared with shared resource groups, dedicated resource groups have the following benefits:
  • Provides subscription and pay-as-you-go dedicated resource groups. You can select a resource group type as required.
  • Allows you to deploy more types of models by using custom processors.
  • Allows you to connect to your VPC so that you can call models in an efficient and lossless manner.

Types

Dedicated resource groups are classified into subscription dedicated resource groups and pay-as-you-go dedicated resource groups based on billing methods. In this topic, these two resource group types are referenced as subscription resource groups and pay-as-you-go resource groups for short. The following table describes the differences between the two types of dedicated resource groups.
Type Subscription resource group Pay-as-you-go resource group
Billing method Subscription Pay-as-you-go
Resource group operation Scale out, renew, and temporarily scale out Scale out, scale in, and stop

Create resource groups

  1. Go to the Elastic Algorithm Service page.
    1. Log on to the Machine Learning Platform for AI console.
    2. In the left-side navigation pane, choose Model Deployment > EAS-Model Serving.
  2. Create a resource group.
    You can create a subscription resource group or pay-as-you-go resource group as needed.
    • Create a subscription resource group.
      1. In the upper part of the Elastic Algorithm Service page, click New Resource Group and select New Resource Group-Subscription.
      2. On the buy page, set the parameters as required.
        Parameter Description
        Product Type Select EAS Resource Group(Subscription).
        Region The following regions are supported:
        • China
          • China (Hangzhou)
          • China (Shanghai)
          • China (Beijing)
          • China (Shenzhen)
          • China (Hong Kong)
        • Asia Pacific
          • Singapore
          • Indonesia (Jakarta)
        • Europe and America: Germany (Frankfurt)
        • Middle East and India: India (Mumbai)
        Node Specifications The GPU and CPU specifications that are supported by the system. For more information, see Resource types.
        Note The valid values of the Node Specifications parameter vary in different regions.
        Nodes Valid values: 1 to 1000.
        Order Time Valid values: 1 Month, 2 Months, 3 Months, 4 Months, 5 Months, 6 Months, 7 Months, 8 Months, 9 Months, 10 Months, 11 Months, 1 Year, and 3 Years.
        Auto-renewal Specifies whether to automatically renew the resource group after expiration.
      3. Click Buy Now.
        After payment is complete, the system automatically allocates resources. You can view all created resource groups or edit the resource group names on the Elastic Algorithm Service page.
        Note If the resource group fails to be created due to insufficient resources in the region, the system automatically issues a refund order. You will receive a refund in the same form as your original payment.
    • Create a pay-as-you-go resource group:
      1. On the Elastic Algorithm Service page, click New Resource Group and select New Resource Group-Pay as you go.
      2. On the buy page, set the parameters as required.
        Parameter Description
        Product Type Select EAS Resource Group(Pay as you go).
        Node Specification The GPU and CPU specifications that are supported by the system. For more information, see Resource types.
        Note The valid values of the Node Specification parameter vary in different regions.
        Region The following regions are supported:
        • China (Hangzhou)
        • China (Shanghai)
        • China (Beijing)
        • China (Shenzhen)
        • China (Hong Kong)
        • Singapore
        • Indonesia (Jakarta)
        • India (Mumbai)
        • Germany (Frankfurt)
        Number of Nodes Valid values: 1 to 100
      3. Click Buy Now.

        After payment is complete, the system automatically allocates resources. You can view all created resource groups or edit the resource group names on the Elastic Algorithm Service page.

View the details of a resource group

  1. In the upper-right corner of the Elastic Algorithm Service page, click View All Resource Groups.
  2. On the Resource Group List page, find the resource group that you want to view and click the ID or name in the Resource Group ID/Name column.
  3. On the details page of the resource group, you can perform the following operations.
    Operation Description
    View the ID of the resource group . The ID of a resource group uniquely identifies the resource group. This parameter is required when you use EASCMD to deploy a model. Resource group ID
    Edit the resource group name. Click the Edit icon icon next to the resource group name to edit the resource group name. We recommend that you use a name that represents your core business. This facilitates you select a resource group based on the resource group name when you deploy a model in the console.
    View the Elastic Compute Service (ECS) instances in the resource group. Click the Machine List tab.
    View the services that are deployed in the resource group. Click the Service List tab.

Manage subscription resource groups

  • Scale out a subscription resource group
    This feature allows you to increase the number of ECS instances in a subscription resource group within its validity period. New ECS instances will expire at the same time as existing ECS instances. To scale out a resource group, perform the following operations:
    1. On the Elastic Algorithm Service page, find the subscription resource group that you want to scale out and click Scale-out.
    2. On the Upgrade/Downgrade page, set the Region and Nodes parameters.
    3. Select EAS (Subscription) Terms of Service.
    4. Click Buy Now.
  • Renew a subscription resource group
    This feature allows you to extend the validity period of a resource group. For example, if you renew a subscription resource group for one month, the validity period of the subscription resource group is extended by one month. To renew a subscription resource group, perform the following operations:
    1. On the Elastic Algorithm Service page, find the subscription resource group that you want to renew and click Renew.
    2. On the Renew page, set the Order Time parameter and select EAS (Subscription) Terms of Service. Use the default value for other parameters.
    3. Click Buy Now.
    Note You cannot delete subscription resource groups. To release an unexpired subscription resource group,submit a ticket.
  • Enable the VPC direct connection channel

    This feature allows you to use an Elastic Network Interface (ENI) to connect your subscription resource group to a VPC in the current region. For more information about how to enable this feature, see VPC direct connection channel.

  • Temporarily scale out a subscription resource group
    This feature allows you to purchase temporary ECS instances to increase available resources during peak hours. This can ensure the stability of online services. After the peak hours, you can stop the temporary ECS instances at any time. The following rules are defined for temporary scale-out:
    • Temporary ECS instances are billed in pay-as-you-go mode.
    • After your business demands decrease, you can release the temporary ECS instances at any time. Then, the system stops billing for them.
    • If you do not release a temporary ECS instance, the temporary ECS instance will expire at the same time as the existing subscription ECS instance in the resource group.
    • When you use temporary ECS instances, make sure your account balance is sufficient. This avoids service downtime. For more information about how to handle service downtime, see Manage the validity period of pay-as-you-go resource groups.
    To temporarily scale out a resource group, perform the following operations:
    1. On the Elastic Algorithm Service page, find the subscription resource group that you want to temporarily scale out and click Temporary Scale-out.
    2. In the Purchase Temporary Node dialog box, set the Number of Nodes parameter. Use the default value for other parameters.
    3. Click Confirm Scale-out.
    To release an ECS instance, perform the following operations:
    1. On the details page of the resource group, click the Machine List tab.
    2. In the Purchase Temporary Node dialog box, set the Number of Nodes parameter. Use the default value for other parameters.
    3. Find the temporary ECS instance that you want to release and click Freed in the Operating column.

Manage pay-as-you-go resource groups

  • Scale out a pay-as-you-go resource group
    This feature allows you to increase the number of ECS instances in a specified pay-as-you-go resource group. To scale out a pay-as-you-go resource group, perform the following operations:
    1. On the Elastic Algorithm Service page, find the pay-as-you-go resource group that you want to scale out and click Scale-out.
    2. On the Update page, set the Number of Nodes parameter and select EAS Resource Group (Pay-As-You-Go) Agreement of Service.
    3. Click Activate.
  • Scale in or stop a pay-as-you-go resource group
    This feature allows you to reduce the number of ECS instances in a pay-as-you-go resource group. When the number of ECS instances is reduced to 0, the resource group becomes unavailable. To scale in or stop a pay-as-you-go resource group, perform the following operations:
    1. On the Elastic Algorithm Service page, find the pay-as-you-go resource group that you want to manage and click Scale in/Stop.
    2. If you confirm that you no longer use the ECS instances that you want to release, click I know in the message.
    3. On the Update page, set the Number of Nodes parameter and select EAS Resource Group (Pay-As-You-Go) Agreement of Service.
      Note If you want to stop the pay-as-you-go resource group, set the Number of Nodes parameter to 0.
    4. Click Activate.
  • Delete a pay-as-you-go resource group
    You can delete a stopped pay-as-you-go resource group that you no longer use. To delete a pay-as-you-go resource group, perform the following operations:
    1. On the Elastic Algorithm Service page, find the pay-as-you-go resource group that you want to delete and click Delete. Delete the pay-as-you-go resource group
    2. In the Delete Resource Group message, click OK.
  • Enable the VPC direct connection channel

    This feature allows you to use an ENI to connect your pay-as-you-go resource group to a VPC in the current region. For more information about how to enable this feature, see VPC direct connection channel.

Billing

A subscription resource group is billed upon creation. A pay-as-you-go resource group is billed only when it is created and is in the running state. For more information, see Billing of EAS.

For more information about how to stop or reduce billing for a pay-as-you-go resource group, see the "Scale in or stop a pay-as-you-go resource group" section in this topic.

Manage RAM user permissions

If you want to create, view, and manage resource groups as a RAM user, you must be authorized by your Alibaba Cloud account. For more information about how to authorize a RAM user, see Grant permissions to RAM users.

If you create a resource group as a RAM user, you must have financial permissions. Otherwise, you cannot place orders or make payments. For more information, see Grant permissions to RAM users.

Manage the validity period of subscription resource groups

You can specify that a subscription resource group is automatically renewed, manually renewed, or not renewed upon expiration. To continue to use the resource group after expiration, we recommend that you renew it in advance or enable auto-renewal. Make sure that your account balance is sufficient. If the account balance is insufficient to cover the subscription renewal, the resource group cannot be renewed. The resource group will be stopped if it is not renewed or fails to be renewed upon expiration. If you renew the resource group within 15 calendar days, which are 360 hours, from the expiration date, the system recovers the resource group. If you do not renew the resource group within 15 calendar days, which are 360 hours, from the expiration date, the system terminates the resource group and deletes the resource group and all its data after 15 calendar days.

Manage the validity period of pay-as-you-go resource groups

After a pay-as-you-go resource group is created, the system starts to bill for it. The system generates bills and deducts fees based on the service time of the resource group. If you no longer use the resource group, we recommend that you stop the resource group to stop billing. When you use a pay-as-you-go resource group, make sure that your account balance is not less than the bill amount of the previous billing cycle. If you pay the overdue bill within 24 hours from the overdue date, the resource group can still be available. Otherwise, the system stops the resource group.

If you settle the bill within 15 calendar days, which are 360 hours, from the overdue date, the system recovers the resource group and retains its data. If you do not settle the bill within 15 calendar days, or 360 hours, from the overdue date, the system terminates the resource group and deletes the resource group and all its data after 15 calendar days.

Resource types

EAS allows you to create resource groups by using 14 types of resources. Resources vary with regions. The following table describes the configurations of different resources.
ECS instance type GPU configuration CPU configuration
ecs.c5.6xlarge, which is available only for subscription resource groups N/A 24 cores + 48 GB
ecs.g5.6xlarge, which is available only for subscription resource groups N/A 24 cores + 96 GB
ecs.g6.4xlarge, which is available only for pay-as-you-go resource groups N/A 16 cores + 64 GB
ecs.g6.6xlarge, which is available only for pay-as-you-go resource groups N/A 24 cores + 96 GB
ecs.gn5i-c4g1.xlarge 1 NVIDIA Tesla P4 4 cores + 16 GB
ecs.gn5i-c8g1.2xlarge 1 NVIDIA Tesla P4 8 cores + 32 GB
ecs.gn6i-c4g1.xlarge 1 Tesla T4 4 cores + 15 GB
ecs.gn6i-c8g1.2xlarge 1 Tesla T4 8 cores + 31 GB
ecs.gn6i-c16g1.4xlarge 1 Tesla T4 16 cores + 62 GB
ecs.gn6i-c24g1.6xlarge 1 Tesla T4 24 cores + 93 GB
ecs.gn5-c4g1.xlarge 1 NVIDIA P100 4 cores + 30 GB
ecs.gn5-c8g1.2xlarge 1 NVIDIA P100 8 cores + 60 GB
ecs.gn5-c28g1.7xlarge 1 NVIDIA P100 28 cores + 112 GB
ecs.gn6v-c8g1.2xlarge 1 NVIDIA V100 8 cores + 32 GB