This topic describes the billing rules, overdue payments, and renewal policies of PAI-Rec.
Billable items
Billable item | Description | Billing method |
PAI-Rec instance | PAI-Rec is billed based on instances. In addition to Standard Edition instances, you can purchase the following value-added features:
| |
Associated cloud services | To build a complete recommendation system, you need to use cloud services such as MaxCompute, PAI-EAS, Hologres, OSS, PAI, and Flink in addition to purchasing PAI-Rec instances. The related resource consumption will be billed based on the billing rules of the corresponding cloud services. Note that PAI-TF tasks can be set to use subscription MaxCompute resources; if you use pay-as-you-go PAI-TF tasks, they are billed according to the PAI platform product billing. The PAI platform algorithm binning is billed according to data_analysis billing; pai_etrec and swing are billed according to the default algorithm type. | For information about billing rules, see the following topics:
|
Solution implementation and deployment of advanced features (optional) | If you want Alibaba Cloud engineers to build a recommendation system and customize solutions, contact Alibaba Cloud customer service. After the custom services are complete, you need to pay for the customization. | Contact your account manager for consultation. |
Billing methods
PAI-Rec supports the subscription (prepaid) billing method. You pay for a PAI-Rec instance of the appropriate specification based on your estimated business volume. The payment is made before you use the service. From the date of purchase, the instance is valid for one year. Instances that are not used after the validity period expires will automatically become invalid. You can purchase instances as needed.
The prices of the recommendation engine vary by region due to different resource costs.
Area | Region | Standard Edition price | Recommendation solution customization (value-added feature) | Operations tools (value-added feature) |
Asia Pacific |
| USD 752/month | USD 301/month | USD 151/month |
| USD 1,204/month | |||
Europe and Americas |
|
Overdue payments
Cause
The balance of your account is insufficient.
Service suspension
If a node expires or fails to be renewed, the node is released. After the node is released, it stops providing services and the PAI-Rec services that use the node enter the waiting state.
If you renew the node within 15 calendar days (360 hours) after the expiration date, the node is automatically recovered.
If you do not renew the node within 15 calendar days (360 hours) after the expiration date, the node is permanently deleted.
During the service period of the PAI-Rec instance, to avoid restrictions on the use of the instance (including the consumption of cloud resources such as MaxCompute and PAI), you need to ensure that your Alibaba Cloud account has sufficient balance.
View the overdue amount
Log on to the Expenses and Costs console to view the overdue amount.
Renewal policies
You can renew your instance using one of the following methods:
Auto-renewal
If you do not want to manually renew your instance each time, you can select Auto-renewal when you purchase a PAI-Rec instance.
NoteIf you enable auto-renewal, the system first attempts to deduct fees at 08:00 on the fifteenth day before your PAI-Rec instance expires. If you plan to pay the fee using a credit card, make sure that your credit card has sufficient balance. The system attempts to deduct the fees for auto-renewal each day during the fifteen days before the instance expires until the payment is completed. You must manually renew your instance if it is about to expire the next day.
If you manually renew an instance before the fees are automatically deducted, the system automatically renews the instance before it expires.
Manual renewal
Before the instance expires or within fifteen days after expiration, log on to the PAI-Rec console, and on the Basic Information page, select the target instance and click Renewal.
Appendix: PAI-Rec cost estimation reference (Standard Edition instance + value-added features)
This cost estimation is for reference only. You can view your bill for the actual prices.
The complexity of recommendation solutions leads to significant cost differences. Factors include the number of items and users, the use of vector retrieval, item cold-start algorithms, complex ranking models, and online learning. To reduce costs, you can take the following measures as needed: use EAS instances with auto-scaling, use prepaid MaxCompute, regularly clean up unused data in Hologres or OSS, and use incremental training instead of full training.
The following table outlines the estimated fees for a complete recommendation system (including offline training and online services) based on daily active users (DAU), for reference only.
Business scale | Estimated resource consumption (price/month) |
Fewer than 50,000 DAU | USD 3,000 to USD 6,000 |
50,000 to 100,000 DAU | USD 5,300 to USD 10,500 |
100,000 to 200,000 DAU | USD 10,500 to USD 22,600 |
200,000 to 500,000 DAU | USD 22,600 to USD 45,000 |
500,000 to 2,000,000 DAU | USD 52,600 to USD 105,000 |