All Products
Search

This Product

    Function Compute:Quotas and limits

    Document Center

    Function Compute:Quotas and limits

    Last Updated:Aug 22, 2026

    Function Compute enforces resource quotas to prevent runaway costs caused by misconfigurations or code errors such as recursive invocations and infinite loops. This page lists all quotas and limits, explains which ones are adjustable, and shows you how to request an increase.

    How quota management works

    Function Compute uses two systems to manage quotas:

    • Quota Center — For quotas you can view and increase through the console. Go to Quota Center to manage these quotas without submitting a ticket.

    • Fixed limits — For resources not in Quota Center. Check the Adjustable column in each table below. If a limit is adjustable, submit a ticket or join the DingTalk user group (ID: 64970014484) to request an increase.

    Important

    Quota Center lets you request increases within a predefined range. Some quotas can be increased beyond that range by submitting a ticket. The actual quotas are subject to Quota Center. For quota increase requests to be processed efficiently, provide a detailed business justification in the Reason field. Otherwise, the request is rejected.

    Manage quotas in Quota Center

    Function Compute is integrated with Alibaba Cloud Quota Center. The following quotas are managed in Quota Center.

    Quota Description Default limit
    Maximum number of instances Per region 300
    Maximum size of a CPU image Compressed image for non-GPU function deployments 10 GB
    Maximum size of a GPU image Compressed image for GPU function deployments 30 GB
    Maximum number of layers per function — 5
    Maximum number of triggers per function — 50
    Code package size (console, developer tools, or OSS) ZIP or JAR file 500 MB (selected regions) / 100 MB (other regions)
    The 500 MB code package limit applies to the following regions: China (Hangzhou), China (Shanghai), China (Shenzhen), China (Beijing), China (Zhangjiakou), China (Hohhot), China (Hong Kong), Singapore, Japan (Tokyo), US (Virginia), and Germany (Frankfurt). In all other regions, the limit is 100 MB.

    View and request quota increases

    1. Log on to Quota Center.

    2. In the left navigation pane, choose Product List > General Quota.

    3. On the General Quota Products page, select Elastic Computing from the Product Category drop-down list, then click Function Compute.

    4. On the General Quotas page, you can:

      • View quotas: See the description and current value for each quota.

      • Request an increase: If the quota is adjustable, click Apply in the Actions column. For details, see Create an application to increase a quota.

      • View application history: Click Application History in the Actions column.

    Limits

    The following limits apply to resources not managed in Quota Center. Where a limit is adjustable, submit a ticket or join the DingTalk user group (ID: 64970014484) to request an increase.

    Function resource limits

    Resource Limit Adjustable
    NAS mount targets per function 5 No
    OSS mount targets per function 5 No
    Concurrency per instance 1–200 No

    Function runtime limits

    Limit CPU instances GPU-accelerated instances Adjustable
    Temporary disk space 10 GB 60 GB No
    File descriptors 100,000 100,000 No
    Total processes and threads 1,024 1,024 No
    Maximum memory per function 32 GB 32 GB No
    Maximum function runtime 86,400 s 86,400 s No
    Maximum Initializer hook runtime 300 s 300 s No
    Maximum PreStop hook runtime 90 s 90 s Yes
    Synchronous invocation payload 32 MB 32 MB No
    Asynchronous invocation payload 128 KB 128 KB Yes, up to 256 KB
    Bandwidth 1–5 Gbit/s 1–5 Gbit/s No
    Single log entry size 32 KB 32 KB No
    Code package size (SDK or API, ZIP, Base64-encoded) 100 MB N/A Yes
    ZIP packages uploaded via SDK must be Base64-encoded, which increases the package size. When creating or updating a function, the request body includes the code package and function configuration. Make sure the total size of the Base64-encoded package and all other content in the request body is less than 100 MB.

    Instance delivery speed

    Item Limit Adjustable
    Instance delivery speed 300 instances per minute Yes

    Trigger limits

    Resource Limit Adjustable
    Native OSS triggers per bucket 10 No
    If 10 native OSS triggers per bucket is insufficient, switch to EventBridge-based OSS triggers, which support up to 50 triggers per bucket. To go beyond 50, request an increase in the Number of event rules for a single event bus quota in EventBridge.

    Layer limits

    Limit CPU instances GPU-accelerated instances Adjustable
    Layer size 500 MB 500 MB Yes
    Number of layer versions 100 100 No
    Total size of all layers per function 2 GB 2 GB Yes

    GPU resource limits (per account per region)

    The on-demand GPU limit is included in the total GPU limit.
    GPU type Total limit On-demand limit Adjustable
    Tesla GPUs 10 GPU cards 3 GPU cards Yes
    Ada.1 GPUs 3 GPU cards 1 GPU card —
    Ampere GPUs 3 GPU cards 3 GPU cards —
    Other GPU types 0 0 —

    Limits on accessing other Alibaba Cloud services or resources

    If your region supports VPC access for Function Compute, the following network restrictions apply when calling other Alibaba Cloud services or resources from your functions:

    • ECS instances in a classic network: You cannot use the private IP address of a classic-network ECS instance to access resources on the server, such as web services or file systems. Use the public IP address instead, or migrate the resources to a VPC.

    • RDS instances in a classic network: use the public IP address instead of the private IP address, or migrate the instance to a VPC.

    • Alibaba Cloud services: use the VPC endpoint or public endpoint instead of the private endpoint.

    Compute node limits per account per region

    The scheduling service provides compute nodes with logical cores and memory. Resource pooling provides a burst capacity equivalent to 50% of the total vCPU and memory limits for your region. After pooled resources are exhausted, the scheduler provisions additional resources at the scale-out rate shown in the table below. Exceeding the quota triggers a throttling error. For example, if you need 600 vCPUs and 1,200 GB of memory, the first 300 vCPUs and 600 GB are provisioned quickly from the pooled burst capacity, and the remaining 300 vCPUs and 600 GB are provisioned at the scale-out rate.

    Example (China (Hangzhou), 300 max instances):

    • Function configured with 1 vCPU and 1,024 MB: 300 concurrent instances consume 300 vCPUs and 300 GB — within the regional limits of 600 vCPUs and 1,200 GB.

    • Function configured with 16 vCPUs and 32,768 MB: 300 concurrent instances consume 4,800 vCPUs and 9,600 GB — exceeds the regional limits. Request a compute node quota increase to run this workload.

    Region vCPU limit Memory limit (GB) Scale-out speed
    China (Hangzhou) 600 1,200 360 vCPUs/min, 720 GB/min
    China (Shanghai) 600 1,200 360 vCPUs/min, 720 GB/min
    China (Beijing) 600 1,200 360 vCPUs/min, 720 GB/min
    China (Zhangjiakou) 600 1,200 360 vCPUs/min, 720 GB/min
    China (Shenzhen) 600 1,200 360 vCPUs/min, 720 GB/min
    China (Hong Kong) 300 600 180 vCPUs/min, 360 GB/min
    Singapore 300 600 180 vCPUs/min, 360 GB/min
    US (Virginia) 300 600 180 vCPUs/min, 360 GB/min
    China (Qingdao) 100 200 60 vCPUs/min, 120 GB/min
    China (Hohhot) 100 200 60 vCPUs/min, 120 GB/min
    China (Chengdu) 100 200 60 vCPUs/min, 120 GB/min
    Japan (Tokyo) 100 200 60 vCPUs/min, 120 GB/min
    Germany (Frankfurt) 100 200 60 vCPUs/min, 120 GB/min
    Malaysia (Kuala Lumpur) 100 200 60 vCPUs/min, 120 GB/min
    Indonesia (Jakarta) 100 200 60 vCPUs/min, 120 GB/min
    US (Silicon Valley) 100 200 60 vCPUs/min, 120 GB/min
    China (Ulanqab) 50 100 30 vCPUs/min, 60 GB/min
    Thailand (Bangkok) 50 100 30 vCPUs/min, 60 GB/min
    UK (London) 50 100 30 vCPUs/min, 60 GB/min
    South Korea (Seoul) 50 100 30 vCPUs/min, 60 GB/min
    SAU (Riyadh - Partner Region) 50 100 30 vCPUs/min, 60 GB/min

    What's next