DataWorks provides two types of shared resource groups: the shared resource group for scheduling and the shared resource group for DataService Studio. Both are provisioned automatically when you activate DataWorks — no separate purchase required. Shared resource groups are multi-tenant: all tenants draw from the same resource pool, which means tasks from different tenants compete for resources during peak hours.
When to use a shared resource group
Use a shared resource group only when all of the following conditions apply to your workload:
Question | Use shared resource group | Use serverless resource group |
How many nodes run concurrently? | Small number, well under the parallel limit | Many nodes, or close to the limit |
Is strict data timeliness required? | No | Yes |
Does the data source use a virtual private cloud (VPC)? | No | Yes |
Is an IP address whitelist configured on the data source? | No | Yes |
Do you need frequent, high-concurrency API calls via DataService Studio? | No | Yes |
If any condition in the right column applies, use a serverless resource group instead.
Limits
The following table summarizes the limits for each shared resource group type.
Resource group type | Parallel node limit | Peak hours | Network constraints |
Scheduling | Up to 40 nodes. During peak hours, actual count may drop below 40. | 00:00–09:00 | No VPC access; no IP address whitelist on data sources; public IP/domain access requires sandbox whitelist configuration |
DataService Studio | N/A (cannot meet frequent, high-concurrency API requirements) | N/A | No VPC access; Internet and classic network accessible |
Data Integration | Up to 5 nodes when idle. Contention may reduce the actual count below 5. | N/A | N/A |
Additional constraints (all shared resource groups)
Memory size is read-only. The formula is:
Number of parallel nodes × 512 MB.You cannot modify the underlying network configuration.
You cannot configure an IP address whitelist on a shared resource group. To associate a resource group with a VPC or configure an IP address whitelist, use a serverless resource group.
What happens when resources are exhausted
When the shared resource pool is at capacity, tasks are queued and may not run at their scheduled time. The parallel node count shown above is the ceiling under ideal conditions — during peak hours (00:00–09:00), available capacity may be lower. You can adjust only the number of parallel nodes; memory allocation scales automatically with it.
Network connectivity
A DataWorks resource group is a group of Alibaba Cloud Elastic Compute Service (ECS) instances. Before running nodes, confirm that the resource group can reach your data sources and that security settings do not block the connection.
Shared resource group for scheduling
Scenario | Supported | Action required |
Data source with no IP address whitelist | Yes | None |
Data source with an IP address whitelist | No | |
Data source in a VPC | No | |
Public IP address or domain | Yes (with configuration) | Add the IP address or domain name and port to the sandbox whitelist on the Workspace page in Management Center in the DataWorks console. If the connection still fails, switch to a serverless resource group. |
Security sandbox: The shared resource group for scheduling uses a security sandbox to restrict access from unknown IP addresses. Add your IP address to the security sandbox whitelist before accessing the resource group. For details, see the Security Settings section of "Create and manage workspaces."
Shared resource group for DataService Studio
Network environment | Accessible |
Internet | Yes |
Classic network | Yes |
VPC | No |
IP address whitelist for data sources
To get the IP addresses of the shared resource group machines and add them to a data source whitelist, see Configure an IP address whitelist. For the complete IP address list, see Whitelist of shared Data Integration resource groups.
Billing
Shared resource groups are included with DataWorks activation at no additional cost. Charges apply to the ECS instances in the shared resource group and the data synchronization threads consumed, billed on a pay-as-you-go basis.
For billing details, see: