The MaxCompute automatic scaling feature provides on-demand scaling and a pay-as-you-go billing method for compute units (CUs). The system detects changes in your business load in real time and automatically adjusts the number of CUs to match your job load. You pay only for the CUs that are automatically scaled. This approach ensures job performance while optimizing costs. This topic describes how to configure and use the MaxCompute automatic scaling feature.
Feature overview
In addition to the subscription model, MaxCompute also supports an elastic reserved model that is billed by the hour. This model lets you create custom time plans and scale resources based on configured rules. To further optimize costs and flexibly match resources to business loads, MaxCompute also provides an automatic scaling model. The main differences between these three resource configuration models are as follows:
Resource type | Usage | Billing method | Scenarios |
You can specify a fixed number of compute CUs and purchase them on a monthly or yearly basis for long-term use. | You pay for the number of resources that you purchase. |
| |
In addition to subscription resources, you can scale resources based on custom CU usage and time-based rules.
| You set elastic reserved CUs for custom rules. Billing is based on the purchased amount and duration. Usage is accumulated and billed hourly. |
| |
Automatic scaling | In addition to subscription resources, the system automatically detects job load fluctuations and scales resources to provide the compute CUs required to match the job load. | You pay for the automatically scaled CUs that you use. Billing is based on the usage amount and duration. Usage is accumulated and billed hourly. |
|
Usage notes
You must purchase subscription compute resources before you can configure and use the automatic scaling feature.
The automatic scaling feature is supported for both batch processing and interactive quotas.
When you enable automatic scaling, you must set a maximum CU value that meets your business needs to control costs. The system automatically scales resources within this limit and bills you for the actual number of automatically scaled CUs used.
The automatic scaling capability of MaxCompute is subject to the availability of real-time resources.
The system makes a best effort to meet the maximum limit that you set for automatic scaling but cannot guarantee that the limit can always be reached. If your business requires guaranteed resources, configure elastic reserved CUs at the same time. These resources are reserved for you according to the time plan that you set. For burst demands that far exceed historical levels, such as during holiday sales promotions, contact us in advance to ensure a sufficient resource supply.
Billing method
The automatic scaling feature uses a pay-as-you-go billing method. You pay only for the compute resources you use, with no upfront payment required.
Bills may be delayed. The final bill reflects the actual charges.
Billing unit: The billing unit for compute resource usage is CU-hour.
Metering method: Billing is based on the number of automatically scaled CUs used while automatic scaling is active. The system monitors the actual usage of automatically scaled CUs at a second-level granularity. At the level-1 quota dimension, usage is aggregated and billed on an hourly basis.
For example, if you use 10 CUs per second for the first 30 minutes (1,800 seconds) of an hour and 20 CUs per second for the next 30 minutes (1,800 seconds), the total usage for that hour is:
(10 CUs × 1,800 seconds + 20 CUs × 1,800 seconds) / 3,600 seconds/hour = 15 CU-hours.Billing formula:
Hourly fee = Automatically scaled CU usage in that hour (Unit: CU-hour) × Price per automatically scaled CU.Price per automatically scaled CU: USD 0.0558 /CU-hour.
View usage: If you enable automatic scaling, you can view the corresponding usage details in Billing Details.
Concepts
Maximum automatically scaled CUs (AutoscaleLimitCU)
The upper limit of elastic CU resources that you set for a quota. If this value is greater than 0, the automatic scaling feature is enabled. The system can automatically scale resources based on the actual load, up to this limit.
Used automatically scaled CUs (AutoscaleUsedCU)
The number of automatically scaled CU resources actually consumed by a quota after automatic scaling is enabled. The system automatically adjusts CU usage based on the job load, and you are billed for this actual usage.
Procedure
Configuration flow
Log on to the MaxCompute console, and select a region in the upper-left corner.
In the navigation pane on the left, choose .
On the Quotas page, click Quota Configuration in the Actions column for a quota.
On the Quota Configuration page, select the Basic Configurations tab, and click Edit Basic Configurations.
Click Add Level-2 Quota or configure AutoScale CUs for an existing secondary quota.
The configuration details are as follows:
Level-1 quota:
By default, automatic scaling is disabled. The default value for Maximum Automatically Scaled CUs (AutoscaleLimitCU) is 0.
You can set the maximum number of automatically scaled CUs in increments of 10. The maximum value is the reserved CU value of the level-1 quota.
After you enable automatic scaling, the quota always occupies the reserved
CU + Elastic Reserved CUresources while the configuration is active. It also uses additional automatically scaled CU resources. The number of automatically scaled CUs used ranges from (0, AutoscaleLimitCU] to meet job load demands.
Level-2 quota:
If automatic scaling is enabled for a level-1 quota, it is enabled by default for all its level-2 quotas.
You can set the maximum automatically scaled CUs for each level-2 quota individually. For batch processing quota types, the value can be set in increments of 10 CUs. For interactive quota types, the value can be set in increments of 25 CUs. The value must be within the range of [0, AutoscaleLimitCU of the level-1 quota].
All level-2 quotas share the automatically scaled resources, which the system allocates based on the load of each level-2 quota. The total number of automatically scaled CUs used by all level-2 quotas under a level-1 quota is equal to the number of automatically scaled CUs used by the level-1 quota. This total does not exceed the maximum number of automatically scaled CUs set for the level-1 quota.
For more information about other quota parameters, see Quota management.
The automatic scaling feature can be used with time plans. On the Quota Scaling Configuration page, you can set multiple resource configuration plans and schedule them using Time-based Management.
Resource monitoring
After you enable automatic scaling, you can use resource monitoring to view the usage of automatically scaled CUs for each quota.
Log on to the MaxCompute console, and select a region in the upper-left corner.
In the navigation pane on the left, choose .
On the Resource Observation page, click the Computing Resources tab.
From the Select Quota drop-down list, select the quota that you want to view.
You can view the maximum and used automatically scaled CUs at a second-level granularity for each level-1 and level-2 quota.