If a large number of nodes must be concurrently run, you need exclusive computing resources to make sure that nodes are run as scheduled. In this case, you can use exclusive resource groups for scheduling. This topic describes the details of exclusive resource groups for scheduling.

Limits

  • Exclusive resource groups for scheduling do not support Machine Learning experiment nodes. For more information, see Create a Machine Learning (PAI) node.
  • Assume that you want to change the resource group on which nodes other than MaxCompute nodes are run. Make sure that the compute engine instances of such nodes can access the exclusive resource group for scheduling that you want to use instead.
  • Exclusive resources for scheduling in a single order can be used in only one exclusive resource group for scheduling. The resources cannot be split among multiple exclusive resource groups for scheduling. After you pay for the order, you can increase the number of resources of the same specifications in the order to scale out the resources.

Scenarios

  • The resources in exclusive resource groups can be scheduled at any time to ensure the data output of nodes. We recommend that you use exclusive resource groups for production nodes.
  • You must use exclusive resource groups in scenarios where a large number of nodes are run and data output is required as early as possible.
  • If nodes must connect to the Internet or a virtual private cloud (VPC), use exclusive resource groups.
  • If you need to configure a whitelist for nodes to use external services, use exclusive resource groups.
  • If your VPC is connected to a data center and your nodes need to connect to the data center, bind exclusive resource groups to the VPC.

Billing method

Exclusive resource groups for scheduling are billed in subscription mode. You can choose appropriate specifications for billing as needed. For more information, see Performance metrics and billing standards of exclusive resource groups for scheduling.

Network connectivity

A DataWorks resource group is a group of Alibaba Cloud Elastic Compute Service (ECS) instances. To run a node, such as a data integration or development node, you must make sure that resource groups and data sources are connected. In addition, you must make sure that special security settings, such as whitelists, do not affect the connections between resource groups and data sources.

  • Network connectivity
    • The data store is deployed on the Internet. In this case, the exclusive resource group can directly access the data store.
    • The data store is deployed in a VPC.
      • Assume that the data store and the exclusive resource group reside in the same region. We recommend that you bind the exclusive resource group to the VPC where the data store resides and then to a vSwitch that resides in the VPC. Then, the system automatically adds a route for the exclusive resource group. This way, the exclusive resource group can access the data store.
      • Assume that the data store and the exclusive resource group reside not in the same region. You must connect the VPC where the exclusive resource group resides to the VPC where the data store resides by using Express Connect circuits or VPN gateways.
    • The data store is deployed in a data center. In this case, you can connect the data center to the VPC where the exclusive resource group resides by using Express Connect circuits or VPN gateways.
    • The data store is deployed on the classic network. Exclusive resource groups are deployed in VPCs of Alibaba Cloud. If your data store is deployed on the classic network of Alibaba Cloud, the exclusive resource groups cannot access the data store. In this case, we recommend that you bind the data store to a VPC.
  • Whitelist configuration. Assume that a whitelist is configured for the data store to allow only specified IP addresses to access the data store. You must add the elastic IP address (EIP) of the resource group and the Elastic Network Interface (ENI) IP address of the VPC to the whitelist. For more information, see Obtain the IP address and CIDR block of an exclusive resource group for Data Integration.

Other instructions

  • Scale-out
    You can scale out the exclusive resources for scheduling in each order based on your business needs. You need to pay only for the added resources from the time when you scale out the resources to the time when the original order expires.
    Note
    • Scale-out increases the number of resources but does not upgrade their CPU and memory configurations.
    • Scale-out allows you to add only resources of the same specifications as the purchased resources in the current order.
    • After you scale out the resources in an order, the new configuration takes effect within 20 minutes after you pay for the order.
  • Renewal

    You can renew the exclusive resources for scheduling in an order before the order expires or within 30 days after the order expires. If you do not renew your subscription within 15 days after expiration, the services for resources in the order are suspended immediately after this 15-day period.

  • Expiration
    • Service suspension

      By default, the system sends expiration notifications to the mobile number and email address that are bound to your Alibaba Cloud account eight days, 12 days, and 14 days before your order for the exclusive resource group for scheduling expires. If you do not renew your subscription within 15 days after expiration, the services for resources in the order are suspended immediately after this 15-day period.

    • Resource release

      After the exclusive resource group for scheduling expires, it is retained for 15 days. If you do not renew your subscription within 15 days, the exclusive resource group for scheduling is released after this 15-day period.

      The system sends release notifications to the mobile number and email address that are bound to your Alibaba Cloud account one day before the exclusive resource group for scheduling is released.