Feature Overview
Time-based elasticity (also known as time-based scaling) is an intelligent resource scheduling feature provided by the DLC standard engine. It allows users to configure multiple scaling policies for their yearly/monthly subscription Spark standard engine based on time periods. During peak business hours, it automatically scales out to ensure performance, and during off-peak hours, it automatically scales in to reduce costs.
Compared to the "passive response" of load-based elasticity, time-based elasticity can proactively anticipate business peaks and actively schedule resources. It is suitable for batch processing job scenarios with regular peak and valley characteristics.
Note:
Time-based elasticity and load-based elasticity are independent of each other and can be configured simultaneously. When time-based elasticity rules are active, the system prioritizes their execution. During inactive periods, the system resumes load-based elasticity rules.
Use Limits
Applicable Engines: This feature only supports the yearly/monthly subscription Spark standard engine. It does not support pay-as-you-go engines.
Currently, you cannot set a usage limit for resource group resources individually during elastic time periods.
Elastic Resource Limit: The elastic resource limit for a single Spark standard engine cannot exceed the number of yearly/monthly subscription resources (for example, an engine with 128CU can have a maximum of 128CU of elastic resources). If you need a higher elastic limit, please submit a ticket to contact us. After the scheduled time period elapses, if jobs or SQL-only analysis resource groups are still occupying resources, the system cannot automatically scale in. If you have SQL-only analysis resource group scenarios, we recommend that you enable automatic start and stop.
Operation Steps
Entering the Configuration Portal
Log in to the DLC console and go to the time-based elasticity configuration page by following the path below: Standard Engine > Select the target engine (yearly/monthly subscription Spark) > Operations > More > Specifications Configuration > Elasticity Rules > Enable Scheduled Elasticity.
Configuring the Execution Type
Time-based elasticity supports two execution types: run once and recurring execution.
Executing Once
This applies to one-time large-scale events on specific dates, such as major promotions or holiday business peaks.
Description of configuration items:
|
Time zone | Users can self-select the time zone that takes effect during the elastic time period (supported only in the "Execute Once" mode). |
Effective time period | The start and end time for scheduled elasticity, with a total duration not exceeding 24 hours (supported only in the "Execute Once" mode). |
Maximum scaling value | The maximum scaling value allowed during the elastic time period. |
Repeating Execution
This applies to periodic batch processing jobs with regular peak and valley characteristics, such as daily ETL or weekly reporting tasks.
Multiple scaling plans can be configured under the same execution frequency, and the planned time periods cannot overlap.
Scaling plans are executed in chronological order.
Supported execution frequencies:
Daily
The total duration of the effective period cannot exceed 24 hours.
If the end time is earlier than the start time, the execution is considered as cross-day execution.
Weekly
You can select specific weekdays with multiple selections supported.
Monthly
You can select specific dates with multiple selections supported.
The End of Month option indicates that the execution is performed on the last day of a month (regardless of the number of days in the month).