All Products
Search
Document Center

Platform For AI:EAS resource groups

Last Updated:Apr 03, 2026

EAS provides dedicated resource groups for purchasing and managing dedicated resources, and virtual resource groups for deploying services across multiple resource types such as public resources, dedicated resources, and Lingjun resources.

Dedicated resource groups

Resource types

Dedicated resources support Subscription and Pay-as-you-go billing.

Type

Billing method

Resource group management

Supported node types

Subscription resources

Subscription

Purchase nodes, renew Subscription nodes, delete Pay-as-you-go nodes, and delete resource groups. For more information, see Manage nodes in a dedicated resource group.

For supported node types, go to the EAS Subscription Dedicated Nodes page.

Pay-as-you-go resources

Pay-as-you-go

For supported node types, go to the EAS Pay-as-you-go Dedicated Nodes page.

Important

Pay-as-you-go dedicated resources start billing immediately upon purchase, regardless of service deployment. Delete unused Pay-as-you-go nodes promptly to avoid unnecessary costs.

Create a dedicated resource group

  1. Go to the AI Computing Resources > Resource Pools page and select the EAS Inference Resources tab.

  2. On the Dedicated Resource Group tab, click Create Resource Group and select one of the following options:

    • Create Resource Group: Creates only the resource group. Purchase nodes later from the resource group details page.

    • Create Resource (Subscription) or Create Resource (Pay by Usage): Opens the purchase page where you purchase nodes and create a resource group at the same time.

  3. On the purchase page, configure the parameters described in the following table.

    Parameter

    Description

    Billing Method

    Select Subscription or Pay-as-you-go.

    Region and Zone

    The region and availability zone for the resource group.

    Node Type

    GPU and CPU specifications. For supported node types, see EAS Subscription Dedicated Nodes or EAS Pay-as-you-go Dedicated Nodes.

    Note

    The supported node types vary by region.

    System Disk

    Each node includes 200 GB of free system disk capacity. The storage capacity can be increased from 200 GB to 2,000 GB. For information about charges for extra capacity, see Billing of EAS.

    Note

    All service instances share system disk space. If disk space is insufficient, the system evicts instances that consume excessive space. To read from and write to a local disk directory during runtime, see Mount storage.

    Purchase Option

    Select Create Resource Group or Existing Resource Group.

    If you select Existing Resource Group, you must configure the Resource Group ID parameter.

    Resource Group ID

    Select an existing resource group from the list.

    Quantity

    Valid values: 1 to 9999.

    Subscription Duration

    Valid values: 1 Month, 2 Months, 3 Months, 6 Months, and 1 Year. Applies only to Subscription.

    Auto-renewal

    Automatically renews the subscription upon expiration. Applies only to Subscription.

  1. Click Buy Now.

    After payment, the system automatically provisions the resources. View all resource groups or modify their names on the Resource Group tab.

    Note

    (Applies only to Subscription) If resource group creation fails (for example, due to insufficient resources), the system automatically issues a refund to your original payment method.

Manage nodes in a dedicated resource group

Purchase nodes

  1. On the Resource Group tab, select the target resource group.

  2. Click Purchase Node to purchase Subscription or Pay-as-you-go nodes. For configuration details, see Create a dedicated resource group.

Renew nodes

(For Subscription nodes only)

Extends the usage duration of a node. For example, a one-month renewal extends the expiration date by one month.

  1. On the details page of the dedicated resource group, go to the Nodes tab. In the Actions column of a Subscription node, click the 更多 icon and select Renew.

  2. On the Renew page, configure the Subscription Duration and click Buy Now.

Delete nodes from a dedicated resource group

Pay-as-you-go nodes: On the details page of the dedicated resource group, click Delete in the Actions column for the target node.

Subscription nodes: See EAS FAQ.

View node resource usage

On the Resource Group tab, click the name of the target resource group. On the Nodes tab, view the used/total CPU, GPU, and memory for each node.

Stop scheduling

When a node is in the Running state, click Stop Scheduling in the Actions column to make the node unavailable for scheduling. Click Restart Scheduling to resume scheduling.

Note

After scheduling is stopped, services already running on the node continue to run. If a service restarts, it is redeployed to other available nodes in the resource group.

Clear a node

Click Clear Node. This stops and redeploys service instances on the node to other available nodes in the resource group.

Important

Clearing a node terminates all running tasks on it. Perform this action with caution.

Virtual resource groups

An EAS virtual resource group logically groups various resource types, such as public resources, resource quotas, and EAS dedicated resources. This enables unified service deployment and scheduling across multiple resource types within a single service.

Create a virtual resource group

  1. Go to the AI Computing Resources > Resource Pools page and select the EAS Inference Resources tab.

  2. On the Virtual Resource Group tab, click Create Virtual Resource Group. Public resource groups support spot instance selection. Configure multiple resource types as needed and set the scheduling priority for each resource type (1 = lowest, 9 = highest).

    image

Use a virtual resource group

When deploying a service, in the Resource Information section, select EAS Resource Group for Resource Type. Then, from the Resource Group drop-down list, select Virtual Resource Group.

image

Related documentation

FAQ

  1. Can I deploy multiple models in a single resource group?

    Yes. You can deploy multiple model services in a single resource group, provided it has sufficient resources.

  2. Can surplus EAS inference resources be used for DLC distributed training?

    No. EAS dedicated resources support only service deployment, not DLC distributed training.