MaxCompute:Compute resources - Quota management

Basic configuration

This involves adding or deleting level-2 quotas for cloud data warehouse resource types and configuring basic parameters. These parameters include exclusive use, CU limit per job, priority, non-reserved CUs, and the scheduling policy. It also includes configuring reserved CUs and elastic reserved CUs. This configuration applies only to the current configuration plan. If you do not have time-based scaling scenarios, this configuration is sufficient for quota management.

Scaling configuration

Quota scaling configuration, also known as time-based scaling, lets you set different minimum and maximum reserved CUs or elastic reserved CUs for a quota at different times.

Configuration procedure

1. Log on to the MaxCompute console and select a region in the top-left corner.

2. In the navigation pane on the left, choose Manage Configurations > Quotas.

3. On the Quotas page, find the target quota and click Quota Configuration in the Actions column.

   You can configure only quotas for which the Billing Method is Subscription.

4. Perform basic quota configuration.

   1. On the Quota Configuration page, click the Basic Configurations tab, and then click Edit Basic Configurations.

      The following table describes the parameters.

      Parameter	Description
      Quota Name	A custom name. The name must start with a letter and can contain letters, digits, and underscores (_).
      Type	Batch: A quota used to run batch jobs. Interactive: A quota dedicated to running query acceleration (MCQA) jobs. You do not need to configure it as the default compute quota for a project. If the default compute quota for a project is a subscription batch processing quota, jobs that hit query acceleration are automatically scheduled to an interactive quota for execution. If there are multiple interactive quotas, MCQA jobs are scheduled to the earliest created interactive quota by default. Therefore, you need to use quota rules to schedule MCQA jobs initiated by different projects to different interactive quotas. For more information about quota rules, see Quota rules.
      Reserved CUs [minCU,maxCU]	The monthly reserved CUs. minCU is the guaranteed value, and maxCU is the upper limit. The reserved CUs of a level-1 quota is the number of CUs purchased on a subscription basis (minCU=maxCU). To scale out or scale in, you must upgrade or downgrade the quota. This parameter cannot be edited here. If maxCU > minCU, the level-2 quotas are in shared mode. If the requested CUs exceed the min value, idle and non-exclusive CUs from other quotas will be preempted. The sum of the minCU values of all level-2 quotas equals the minCU value of the level-1 quota. The minCU of a level-2 quota must be less than or equal to its maxCU, which must be less than or equal to the minCU purchased for the level-1 quota. The minCU and maxCU of the default level-2 quota cannot be edited: minCU of the default level-2 quota = [minCU of the level-1 quota] - [Sum of the minCU values of custom level-2 quotas]; The maxCU value is the same as the maxCU value of the level-1 quota. The minCU of a level-2 quota cannot be set to 0 for now. For a new quota This configuration applies only to the currently active configuration plan. If multiple scaling configuration plans exist, the new quota is added to all of them after it is created. In plans other than the currently active one, the default reserved minCU is 1 (deducted from the default subscription level-2 quota). Therefore, to adjust the number of CUs, edit the configuration plans on the Scaling Configuration tab.
      Elastically Reserved CUs	The number of elastic CUs reserved on an hourly basis. For more information about the billing of elastic reserved CUs, see Compute costs (hourly billing). The elastic reserved CUs for a level-1 quota is the number of CUs you need to reserve. The default value is 0. The maximum value is the number of subscription reserved CUs purchased for the quota, and cannot exceed 10,000 CUs. This is also subject to inventory, which means an inventory check is performed during scale-out configuration. The scale-out capacity must be at least 50 CUs or a multiple of 50. The elastic reserved CUs for a level-2 quota are defined as follows: The sum of the elastic reserved CUs of all level-2 quotas equals the elastic reserved CUs of the level-1 quota. The elastic reserved CUs of the default level-2 quota cannot be edited. It is calculated by subtracting the sum of the elastic reserved CUs of custom level-2 quotas from the elastic reserved CUs of the level-1 quota. For a new quota This configuration applies only to the currently active configuration plan. If multiple scaling configuration plans exist, the new quota is added to all of them after it is created. In plans other than the currently active one, the default elastic reserved CUs is 0. Therefore, to adjust the number of CUs, edit the configuration plans on the Scaling Configuration tab.
      Exclusive or Not	Specifies whether to make the quota strongly exclusive. If a quota is exclusive, its CUs cannot be occupied by other quota groups even if they are idle. This parameter can be configured only for level-2 quotas. If the level-2 quota type is Interactive, this parameter is set to exclusive by default and cannot be modified. This applies to quota groups that run services such as BI or ALGO. Jobs of this type may be used at any time. You must prevent them from being occupied by other quotas for a long time without being released. When the max value of a quota is greater than its min value, it may occupy other idle quotas.
      Maximum Number of CUs for a Single Job	Sets the upper limit on concurrent CUs for jobs that run in this quota. An empty value indicates no limit. A value of 0 indicates an absolute limit. You can enter a positive integer to set a reasonable limit. If the level-2 quota type is Interactive, this parameter is unavailable by default. This prevents a single job from occupying many CUs for a long time, which would cause other jobs to wait for resources. You can also set a limit at the job level by running the set odps.task.max.concurrent.cu=<CU_amount>; command before the code to be executed. The job-level configuration has a higher priority than the quota-level configuration.
      Enable Priority	The priority feature. For more information, see Job priority. Jobs that run on this quota are executed with priority. This also takes effect if priority is enabled at the project level. If the level-2 quota type is Interactive, this parameter is unavailable by default.
      Non-reserved CUs	The number of monthly non-reserved CUs. The non-reserved CUs of a level-1 quota is the number of non-reserved CUs purchased on a subscription basis. To scale out or scale in, you must upgrade or downgrade the quota. Scale-out is not possible due to current inventory shortages. This parameter cannot be edited here. The non-reserved CUs of a level-2 quota are defined as follows: The sum of the non-reserved CUs of all level-2 quotas equals the non-reserved CUs of the level-1 quota. The non-reserved CUs of the default level-2 quota cannot be edited. It is calculated by subtracting the sum of the non-reserved CUs of custom level-2 quotas from the non-reserved CUs of the level-1 quota.
      Scheduling Policy	MaxCompute level-2 quota scheduling policies include FAIR and FIFO. FAIR (Fair scheduling) If jobs have the same priority, resources are evenly allocated to all jobs submitted at the same time. If jobs have different priorities, resources are first evenly allocated to higher-priority jobs. If resources remain, they are then evenly allocated to lower-priority jobs. FIFO (First in, first out) If jobs have the same priority, resources are allocated to the job that was submitted first. If jobs have different priorities, resources are allocated to the higher-priority job, even if it was submitted later than a lower-priority job. Select an appropriate scheduling policy as needed. You cannot set a scheduling policy for interactive level-2 quotas. For more information, see Compute resources - Quota scheduling policies. Important Starting from May 24, 2023, the default scheduling policy for new level-2 quotas in the China (Hangzhou), China (Shanghai), China (Beijing), China (Zhangjiakou), China (Ulanqab), China (Shenzhen), and China (Chengdu) regions is FIFO. For level-2 quotas created in these regions before this date, the default scheduling policy is FAIR.

   2. Delete a quota.

      On the Quota Configuration page, click the Basic Configurations tab, and then find the target quota and click Delete in the Actions column.

      After a level-2 quota is deleted, it is removed from all resource configuration plans. Its allocated minimum CUs and elastic reserved CUs are returned to the default level-2 quota. To make adjustments, you can edit and update each configuration plan on the Scaling Configuration tab.

5. Configure quota scaling.

   After you purchase a subscription MaxCompute project, each level-1 quota has a default resource configuration plan named Default.

   Each resource configuration plan includes the reserved CU and elastic reserved CU configurations for a level-1 quota and all its level-2 quotas. You can use multiple configuration plans with time-based management for scenarios where different configurations are needed at different times of the day. For example, if the CU requirements for a level-2 quota differ between the 00:00–08:00 and 08:00–24:00 time periods, you can create two quota configuration plans and use time-based management to schedule them.

   Note

   You can add up to 48 resource configuration plans. Each plan can have a different configuration to be enabled at different times of the day.

   1. On the Quota Configuration page, click the Scaling Configuration tab.

   2. On the Scaling Configuration tab, click Add Configuration Plan. In the Add Configuration Plan dialog box, enter a Configuration Plan Name and configure Reserved CUs [minCU,maxCU] and Elastically Reserved CUs for the level-2 quota.

   3. Click OK to add the plan.

   4. On the Scaling Configuration tab, find the target configuration plan and click Apply Immediately in the Actions column to apply it immediately. You can also use an existing plan when you configure Scheduled Scaling Management.

      You can click Apply Immediately for any configuration plan. If a time-based management schedule is set, the plan is overwritten by the next scheduled plan. Therefore, if you want to manually activate a specific plan and keep it active, you must clear all other time-based settings. This ensures that your action takes effect immediately and is not overwritten.

6. (Optional) Perform other operations on resource configuration plans as follows:

   • View a configuration plan

     On the Scaling Configuration tab, click the icon to the left of Configuration Plan Name to expand the plan details and view the information for each plan.

   • Edit a configuration plan

     On the Scaling Configuration tab, find the target configuration plan and click Edit in the Actions column to update it.

     If you edit the currently active plan, you must also click Apply Immediately in the Actions column to apply the changes.

   • Delete a configuration plan

     On the Scaling Configuration tab, find the target configuration plan and click Delete in the Actions column to delete it.

     The Default plan and the currently active plan cannot be deleted.

   • Clone a configuration plan

     On the Scaling Configuration tab, find the target configuration plan and click Clone in the Actions column to create a copy of the plan.

7. Configure time-based management.

   You can set different quota configuration plans to take effect at different times of the day. This allows for time-based management of quota configurations.

   Note

   Time-based scheduling may have a delay of up to 5 minutes.

8. On the Scaling Configuration tab, click Edit Time Plan in the Scheduled Scaling Management area.

9. Click Add Effective Period and select an Effective Start Time and a Configuration Plan.

   Note

   • Each quota has a single time plan. The default runtime is 00:00, which cannot be modified or deleted. This runtime corresponds to the Default configuration plan, which can be modified. If you do not add more runtimes, the same configuration plan remains active throughout the day.

   • You can enable different configuration plans on the hour or half-hour. A configuration plan can be reused.