The scaling plan feature for subscription serverless resource groups uses a hybrid model of "reserved CU + elastic CU". You can configure scheduled elastic plans to handle business peaks and cyclical workload changes.
Overview
To improve resource utilization and reduce long-term costs, DataWorks offers scaling plans for subscription serverless resource groups, which use a hybrid model of "reserved CU + elastic CU":
Reserved CU: The Reserved computing resources (CU) that are allocated when you purchase a subscription resource group.
elastic CU: Elastic computing resources (CUs) dynamically provisioned by a scaling plan.
You can add elastic CUs to your reserved CUs to handle scenarios such as business peaks, sudden traffic spikes, or periodic workload changes. The system automatically scales out resources during the scheduled period and scales in after the period ends. For more information about the lifecycle states, core mechanisms, and other principles of scaling plans, see Advanced topics for resource group scaling plans.
Limitations
Permissions: You must use an Alibaba Cloud account or a RAM user with the AliyunDataWorksFullAccess permission policy.
Regions: All regions that support serverless resource groups.
Resource group type: Available for subscription serverless resource groups.
Use cases: Elastic CUs apply only to Data Computing and Synchronization Tasks in Data Integration quotas. They do not apply to scenarios such as DataService Studio, Individual Development Environment, or Model Service.
Configure a scaling plan
A scaling plan separates what to scale from when to scale it:
Configuration Plan: Defines the number of Elastic Resource to add and a plan name. This acts as a reusable template and is not executed directly.
Schedule: Defines the trigger time and duration for scaling.
These two components have a one-to-many (1:N) relationship. One Configuration Plan can be reused by multiple Schedules for different business scenarios, improving configuration efficiency. Each Schedule must be bound to a Configuration Plan, which defines what to scale.
You can use the Scheduled Elasticity Overview (CU) chart, the Execution History tab, and the resource group details page to preview the configuration of your scaling plan, monitor the scaling status in real time, and view resource usage and actual elastic CU consumption.
Create Configuration Plan
Log on to the DataWorks console. In the left-side navigation pane, click Resource Groups to go to the Resource Groups page.
Click the target subscription serverless resource group, and then click Scaling Plan in the upper-right corner to go to the Scaling Plan page.
On the Scheduled Elasticity tab, click Create Configuration Plan and enter a name for the plan.
Add Schedule
After creating a Configuration Plan, it is inactive by default with no effective dates or times. To activate the plan for automatic scaling, you must bind it to one or more Schedules.
After you create a scaling plan, it requires time for scheduling and resource initialization. Submit the plan at least 3 hours before its start time to ensure it is processed in time.
On the Scheduled Elasticity tab, click Add Schedule.
In the dialog box that appears, set the effective date and time period. The effective time period must be longer than 1 hour, and the start and end times can be set only in 10-minute increments. Select a previously created Configuration Plan from the drop-down list to bind it to the schedule.
NoteThe effective time ranges of different schedules cannot overlap. Scaling records are retained for a maximum of 60 days.
After you save the settings, a new Schedule record is added to the list. You can view its Effective Date Range, Effective Time Period, Configuration Plan Name, and Status. In the Actions column, you can click Edit to edit the Schedule, or click the
icon to Disabled or Delete the schedule.NoteTo delete a schedule, you must first Disabled it.
To re-enable a disabled schedule, click
> Enable.A schedule can have the following statuses:
Status
Description
Pending
The schedule is waiting to start. This includes the Initializing state.
In Effect
The schedule is in one of the following lifecycle states: Scaling Out, Scale-out Succeeded, or Scaling In. For details on each state, see Advanced topics for resource group scaling plans.
Disabled
The schedule has been manually disabled.
Expired
The schedule's effective period has ended, and the system will no longer execute it.
The system automatically updates the status of the corresponding Configuration Plan. In the list of Configuration Plans, you can view the effective date range and time period associated with the plan.
To delete a Configuration Plan, you must first delete all of its bound schedules. After the Configuration Plan is no longer associated with any schedules, you can go to the Resource Configuration Plans list and click the icon in the Actions column of the target plan to delete it.
Preview plans and monitor execution
Scheduled elasticity overview
On the Scaling Plan page, the Scheduled Elasticity Overview (CU) chart shows the resource distribution over time for a specific date. You can switch the date in the upper-right corner of the chart. The horizontal axis represents the time period, and the vertical axis represents the total number of CUs. Reserved and elastic CUs are color-coded, allowing you to track resource changes and the effects of scaling.
Calendar view
In the upper-right corner of the Scheduled Elasticity tab, click the Calendar View toggle button to view the Scaling Plan by day, week, or month. In the calendar view, you can see the effective times and elastic CU amounts for each plan.
Execution History
On the Scheduled Elasticity tab, switch to the Execution History tab to view details of historical elastic scaling tasks. The details include the Configuration Plan name, actual start time, actual end time, and status.
Column | Description |
Configuration Plan Name | The name of the Configuration Plan Name. This name is defined when you create a Configuration Plan and is used to distinguish different scaling policies. |
Plan Start Time | The scheduled start time for the scale-out operation, as set in the Schedule. This is the baseline time when the system attempts to allocate Elastic Resource. |
Plan End Time | The scheduled end time for the scale-out period, as set in the schedule. This is the baseline time when the system begins to scale in and release Elastic Resource. |
Scale-out CUs | The number of elastic CUs added in this scaling plan. After a successful scale-out, Total CUs = Reserved computing resources (CU) + Value in this column. |
Actual Effective Date | The date on which the Scaling Plan was successfully triggered. |
Actual Start Time | The exact time when the system finished allocating the elastic CUs and the resource group could use the new Elastic Resource. Note The Actual Start Time may be slightly later than the Plan Start Time due to factors such as system scheduling and initialization. |
Actual End Time | The exact time when the system released all elastic CUs, and the resource group returned to its Reserved Specification. |
Status | Normal statuses include Initializing, Scaling Out, Scale-out Timeout, Scale-out Succeeded, Scaling In, Scale-in Timeout, and Scale-in Succeeded. For a detailed explanation of each status, see Advanced topics for resource group scaling plans. If you force stop a scaling plan while it is running, the system updates its status to |
Monitor resource usage
On the Resource group details page, you can view the Elastic Specification and Resource Usage. When you hover the pointer over the Resource Usage chart, a detailed breakdown of four CU types is displayed: Reserved In Use, Reserved - Idle, Elastic - Idle, and Allocated.
Billing
Billing rules
Elastic CUs are billed on a pay-as-you-go basis, which is divided into two phases: scale-out and scale-in. For information about CU unit prices, see CU unit prices for resource groups.
Billing phase | Time range | Billing method |
Scale-out period | Actual start time → Plan end time | You are billed based on the configured elastic CU specification:
|
Scale-in period | Plan end time → Actual end time | During the scale-in phase, you are billed based on the actual usage of elastic CUs. The system samples the total CU usage of the resource group every minute:
After all elastic CUs are released, no further charges are incurred. |
To ensure fair billing, the billing for the scale-out period starts at the Actual Start Time, not the plan start time.
Example
A serverless resource group has a reserved specification of 10 CU. A scaling plan is configured to scale out by 20 CU from 11:00 to 12:00. The actual start time is 11:03:00, and the actual end time is 12:02:50.
1. Scale-out period cost
Actual scale-out duration = 12:00 − 11:03 = 57 minutes.
Cost = 20 CU × (57/60) hours = 19 CU-hour
2. Scale-in period cost
Usage exceeding the 10 reserved CUs is billed on a pay-as-you-go basis:
Sampling time | Total CU usage | Elastic CU usage | CU-hour incurred |
End of 12:01 | 28 CU | 28 − 10 = 18 CU | 18 × 1/60 = 0.3 |
End of 12:02 | 16 CU | 16 − 10 = 6 CU | 6 × 1/60 = 0.1 |
End of 12:03 | ≤ 10 CU | 0 (released) | 0 |
Total cost for the scale-in period: 0.3 + 0.1 = 0.4 CU-hour.
3. Total cost
Total cost = Scale-out cost + Scale-in cost = 19 + 0.4 = 19.4 CU-hour
Quotas and limits
The maximum scale-out size for a single scaling plan is 1,000 CUs and cannot exceed five times the resource group's Reserved Specification. The formula is: elastic CU ≤ min(1000, Reserved Specification × 5).
You can still manually scale the resource group while a scaling plan is running.
A single resource group supports up to 5 active configuration plans and 20 schedules (excluding disabled schedules).
To ensure timely execution, submit a scaling plan at least 3 hours before its start time to allow for system scheduling and resource initialization.
A single resource group can have only one schedule in the In Effect state at any time. Therefore, the effective date and time ranges of all enabled schedules cannot overlap. The system automatically validates and blocks overlapping configurations when you create or edit a schedule.
FAQ
Q: Why did the schedule I created not take effect on the same day?
A: If you submit a scaling plan less than 5 minutes before its start time, or after the start time has passed, the system will not trigger it that day. It will take effect on the next scheduled day.
Q: How can I better avoid scale-out or scale-in timeouts?
A: We recommend the following strategies:
Allow time for plan initialization: Submit plans at least 3 hours before their start time to allow for system initialization.
Set a safety buffer period: Leave a time buffer between consecutive scaling plans to ensure the previous plan's scale-in process completes.
Assess resources and plan accordingly: For unavoidable long-running tasks, you can extend the duration of their scaling plan to cover the entire execution cycle.
Q: Elastic scaling resources support only Data Computing and Data Integration. What should I do if I need to scale out resources for other scenarios, such as DataService Studio, Individual Development Environment, or Model Service?
A: Scaling plans are designed for batch processing scenarios such as Data Computing and Data Integration. For non-batch, long-running scenarios such as DataService Studio, Individual Development Environment, and Model Service, elastic CUs are not applicable. To increase capacity for these scenarios, upgrade the resource group's reserved CUs.