You can purchase a resident resource pool to reserve computing resources of a specific type. Then, you can assign a specific number and type of resident instances to your functions. This method ensures business stability and provides fixed, controllable costs.
Introduction to resident resource pools
Resident resource pools apply only to GPU functions and use a monthly subscription billing method. This helps you reserve scarce GPU resources in advance to ensure your services run smoothly without resource shortages. After you purchase a resident resource pool and attach a specific number and type of resident instances to a function, the function uses these instances to process requests.
Functions that use resident instances cannot use on-demand instances at the same time. The maximum number of requests that can be processed simultaneously is (Number of allocated resident instances) × (Instance concurrency). Requests that exceed this limit are throttled. Requests within the limit receive a real-time response, which completely eliminates cold starts.
Scenarios
You can purchase a resident resource pool if your business has the following characteristics:
High resource utilization: Your business has a continuous and stable demand for resources.
Low latency requirements: You need fast responses to ensure high performance and low latency.
Controllable costs: You want fixed and predictable costs.
Billing
Resident resource pools use a monthly subscription model. You must pay for one month, several months, or a year in advance. For specific prices, see Resident resource pool quotas and pricing. When you use resident instances from a resident resource pool, the function cannot use on-demand instances at the same time. In this case, no extra fees are incurred beyond the subscription fee.
Resident resource pool quotas and pricing
Billable items
The fees for a resident resource pool include GPU card fees and disk fees:
GPU card fees
This includes fees for GPU memory, vCPUs, and memory. Currently, only full cards can be purchased. For more information, see GPU full card specifications, quotas, and monthly pricing.
Disk fees
The disk capacity for a resident resource pool ranges from a minimum of 10 GB to a maximum of 60 GB, with a step size of 10 GB. Disk pricing varies by region:
Region: China (Hangzhou), China (Shanghai), China (Beijing), China (Ulanqab), and China (Shenzhen)
Disk pricing: CNY 1.2/GB/month
Region: US (Virginia), Germany (Frankfurt), and Singapore
Disk pricing: CNY 1.44/GB/month
GPU full card specifications, quotas, and monthly pricing
After you purchase a resident resource pool, the platform converts the total specifications of your purchased resources into available capacity that you can flexibly allocate to functions. You can then create resident instances based on this capacity.
Region | GPU card type | Full card specifications | Monthly pricing |
China (Hangzhou), China (Shanghai), China (Beijing), China (Ulanqab), China (Shenzhen) | Ada series |
| USD 1,026 |
Ada.2 series |
| USD 908 | |
Ada.3 series |
| USD 1,320 | |
Xpu.1 series |
| USD 1,507 | |
Hopper series |
| USD 2,887 | |
US (Virginia) | Ada series |
| USD 1,048 |
Hopper series |
| USD 3,027 | |
Germany (Frankfurt) | Ada series |
| USD 1,576 |
Hopper series |
| USD 4,032 | |
Singapore | Ada series |
| USD 1,289 |
Hopper series |
| USD 3,537 |
For specific prices, see the purchase page.
Manage resident resource pools
Purchase a resident resource pool
Log on to the Function Compute console. In the navigation pane on the left, choose . In the top menu bar, select a region.
On the Resident Resource Pools page, click Purchase Resident Resource Pool. On the purchase page, set Region, GPU Card Type, Number of Cards, Disk Specification per Card, and Subscription Duration. Then, click Buy Now.
View resident resource pools
You can view and filter all purchased resident resource pools in the list. You can also see details about function allocation, remaining quota, and expiration time. This lets you allocate resources efficiently and prevent waste.
Scale out (upgrade) a resident resource pool
Find the target resource pool as described in viewing your resident resource pools and click Scale-out in the corresponding row. Then, follow the on-screen instructions to complete the scale-out. Note that expired resource pools must be renewed and activated before you can scale them out.
Currently, you can only increase the number of cards and the disk specification per card. You cannot change the GPU card type or decrease the number of cards or disk specification. Therefore, scale out with caution to avoid resource waste or improper configuration.
Expiration and renewal policy
Renew your resident resource pool before it expires or its resources are depleted.
Expiration reminders
Message Center sends you renewal reminders 7, 3, and 1 day before your resident resource pool expires.
Effects of expiration
After a resident resource pool expires, it is automatically shut down. The resident instances for the associated functions are released. If you do not promptly change the function's instance type to on-demand, the function becomes unavailable. After being shut down for 90 days, the resource pool is completely released and automatically detached from the associated functions. To resume using it, you must purchase a new resource pool.
Message Center sends you reminders 30, 7, and 1 day before your resource pool is released.
Renewal instructions
Auto-renewal
When you purchase a resident resource pool, you can select the Enable Auto-renewal on Expiration check box to ensure the resource is automatically renewed. You can also log on to Renew, find the target resource in the resource list, and then enable or disable auto-renewal.
Manual renewal
For resources that do not have auto-renewal enabled, you can follow the steps in View resident resource pools, find the target resource pool, and click Renew to the right of the resource pool. Then, follow the on-screen instructions to complete the renewal.
View bills
After a bill is generated for your resources, you can view consumption details and query the corresponding resource bill in Expenses and Costs - Billing Details. For more information, see Query bills.
Unsubscription policy
If you want to terminate a resident resource pool after purchase, Function Compute revokes the resources and issues a refund based on Alibaba Cloud's unsubscription rules. Before you request an unsubscription, make sure you understand the effects and rules. After the unsubscription is successful, you can track the refund status.
Before you unsubscribe
Before you request an unsubscription, detach the resident resource pool from its functions. Log on to the Function Compute console and change the number of resident instances to 0 for all functions attached to the target resource pool.
Effects of unsubscription
An unsubscription cannot be revoked once it is successful. Carefully check the order information to avoid mistakes. If your order was part of a promotion, you cannot participate in the promotion again after unsubscribing. Vouchers and coupons will be voided.
Unsubscription rules
The actual amount paid is the cash amount you paid for the order. This amount does not include any value from vouchers or coupons. The consumed amount is calculated based on your usage of the resident resource pool. The calculation rules vary, as shown in the following table:
Usage duration
Consumed amount calculation rule
< 30 days
(Daily price × Usage duration in days) × 1.2
≥ 30 days
Daily price × Usage duration in days
NoteBefore a renewal order takes effect, if you request an unsubscription for the resident resource pool, you can only unsubscribe from the resource. You cannot unsubscribe from the renewal order.
Request an unsubscription
Log on to the Unsubscribe page with your Alibaba Cloud account. Find the target resource and click Unsubscribe in the Actions column. Then, follow the on-screen instructions to complete the unsubscription. For more information, see Unsubscribe from an effective order.
Before you request an unsubscription, make sure the number of resident instances for the functions attached to the resident resource pool is set to 0.
View unsubscription details
After you successfully request an unsubscription, the refund is typically processed within 2 to 3 business days. The refund only applies to the amount paid in cash or with a stored value card. The portion paid with vouchers or coupons is not refunded. You can track the refund status.
Overdue payments
If your account has overdue payments, you can continue to use the resources in your purchased resident resource pools. However, you cannot purchase new resident resource pools, upgrade existing pools, or renew them. For more information, see Overdue payments.
FAQ
How do I use a resident resource pool?
After you purchase a resident resource pool, you can set the instance type to Resident Instance when you create a GPU function. This attaches the function to the resident resource pool and allocates resident instances. For more information, see Configure resident instances.
Can I scale in a resident resource pool?
To avoid affecting online services, scaling in a resident resource pool is not currently supported. The number of GPU cards and the disk specification per card can only be increased, not decreased.
When purchasing a resident resource pool, how do I correctly set the disk specification per card to meet the required total disk capacity?
The disk specification per card is not directly related to the disk specification in the function configuration. Function Compute converts the total disk specification you purchase into the total available disk capacity that can be allocated to functions. For example, assume Function 1 is configured with a 30 GB disk specification and requires 4 instances. Function 2 is configured with a 60 GB disk specification and requires 2 instances. The total required disk capacity is 30 GB × 4 + 60 GB × 2 = 240 GB. In this case, you need to purchase 6 cards. The total required disk specification is 240 GB. When you purchase the resident resource pool, you can set the disk specification per card to 240 GB ÷ 6 = 40 GB.