All Products
Search
Document Center

Function Compute:Set function quotas

Last Updated:Apr 02, 2026

To prevent excessive function invocations and uncontrolled costs, each Alibaba Cloud account has a default limit of 300 concurrent instances per region. Function Compute also provides function-level quotas to prevent a single function from using too many instances. This protects backend resources and helps avoid unexpected charges.

Use cases

  • Protect the concurrency of critical functions.

    For example, two functions, function-a and function-b, share an account-level quota. function-a is a critical business function, but function-b is prone to excessive invocations that can impact function-a. You can set a quota for function-b to prevent it from using too many instances, ensuring function-a has sufficient instances. You can also set reserved concurrency for function-a to reserve a specific number of instances for it.

  • Protect a downstream service.

    For example, if your Function Compute functions frequently access an RDS database that has limited processing capacity, you can set a quota on the functions that access RDS to prevent them from overwhelming the database.

  • Block a malfunctioning function.

    For example, if you notice a function invoking abnormally, you can set its quota to 0 to block all invocations.

  • Prevent runaway function invocations.

    For example, user actions in a browser or client can be unpredictable. Setting a function-level quota prevents uncontrolled invocations and helps you avoid unexpected charges.

Limitations

  • By default, a single Alibaba Cloud account has a total concurrency limit of 300 instances per region. Your account's actual quota is displayed in the Quota Center. To request a higher quota, submit an application in the Quota Center.

  • The maximum quota for a single function is 90% of the account-level quota for the region. The sum of all function quotas in a region cannot exceed 90% of the account-level quota for the region.

    For example, if the account-level quota for the region is 300, the maximum quota for a single function is 270. If you set the quota for one function to 100, the combined quotas for all other functions in that region cannot exceed 170.

Set the elastic instance quota

Single function

  1. Log on to the Function Compute console. In the left-side navigation pane, choose Function Management > Function.

  2. In the top menu bar, select a region. On the Function page, click the target function.

  3. Click the Scaling configuration tab, and then click Edit in the Function Quota section.

  4. In the Edit Quota panel, set the Elastic Instance Quota and click OK.

Multiple functions

  1. Log on to the Function Compute console. In the left-side navigation pane, choose Elasticity Management > Function Quotas.

  2. On the Function Quotas page, click Create Quota. In the Create Quota panel, select the target functions, set the Elastic Instance Quota, and click OK.

Resource quota locking for GPU functions

After setting the elastic instance quota, your account locks a portion of the total resource quota for the function. Locked resources are reserved exclusively for that function. For example, if you set the elastic instance quota for a GPU function to 2, the system locks a corresponding number of GPU card resources. The locked quota varies by scenario, as shown in the following table.

Note

For the total quotas for different series of GPU cards, see Quotas and limits.

Scenario

Quota locking logic

Example configuration

Locked function quota

Single version, single card type

elastic instance quota × number of cards per instance

Tesla series, single card

Locks 2 Tesla cards

Multiple versions, same card type

Calculate elastic instance quota × number of cards for the version for each version, and use the maximum value.

Version 1: Tesla series, single card

Version 2: Tesla series, 8 cards

Locks 16 Tesla cards

Multiple versions, different card types

For each card type, calculate elastic instance quota × maximum number of cards among all versions of that type independently, locking quotas separately for each card type.

Version 1: Tesla, single card

Version 2: Ada.1 series, single card

Version 3: Ada.1 series, 8 cards

Locks 2 Tesla cards and 16 Ada.1 cards