Billing modes by QPS
Alibaba Cloud Artificial Intelligence Recommendation (AIRec) supports two billing modes by QPS: throttling billing and elastic billing. The throttling billing mode applies only to requests whose traffic does not exceed the recommendation request QPS quota that you specified for your subscription AIRec instance. If the QPS of concurrent requests exceeds the quota, the system returns a message that the quota is exceeded, and no recommendation results are returned. The elastic billing mode provides elastic QPS for your recommendation request QPS quota that you specified for your subscription AIRec instance. The elastic QPS is charged based on the actual usage period. This billing mode provides extra QPS and allows your users to obtain recommendation results even if the QPS of concurrent requests exceeds the quota.
Billing for elastic QPS
The elastic billing mode provides elastic QPS for your recommendation request QPS quota that you specified for your subscription AIRec instance. This billing mode provides extra QPS and allows your users to obtain recommendation results even if the QPS of concurrent requests exceeds the quota.
The elastic QPS quota in the Operations Edition is 0 to 30. You can adjust the quota within this range based on your needs.
By default, the elastic QPS quota in the Standard Edition is equal to the recommendation request QPS quota that you specified for your AIRec instance. If you want to modify the elastic QPS quota, log on to the AIRec console and go to the basic information page of the instance to perform the operation. For example, you set the recommendation request QPS quota to 60. In this case, the default elastic QPS quota is also 60. This indicates that a maximum of 120 concurrent requests are allowed per second. You can also adjust the elastic QPS quota to a value less than 60 but greater than 0 based on a step of 10.
If you provide a special type of business and want to increase the quota of your elastic QPS, submit a ticket.
Evaluation and billable items of elastic QPS
If you enable the elastic billing mode, the fee you need to pay each month consists of the following two parts:
1. The fee for your subscription AIRec instance: If you purchase an instance whose specifications are 1 million users, 1 million items, and 10 QPS, when the instances are deployed in Beijing, Hangzhou and Shenzhen, the price of 10 QPS Standard Edition instances is USD 2130/month; when the instance is deployed in Singapore, the price of 10 QPS Standard Edition instances is USD 3380/month.
2. The fee for the elastic QPS: If the recommendation request QPS quota of your instance is 10, the default elastic QPS quota is also 10, and the total QPS quota is 20. If the number of concurrent requests within an hour is greater than 10 but less than 20, the elastic QPS is charged by hour. When the service is deployed in Beijing, Hangzhou, and Shenzhen, the unit price of elastic QPS is 0.56 USD/unit/hour. When the service is deployed in Singapore, the unit price of elastic QPS is 0.90 USD/unit/hour.
The following table provides an example. For example, the instance is deployed in Singapore:
Usage period | Maximum QPS of actual concurrent requests | Use elastic QPS | Maximum elastic QPS | Price |
|---|---|---|---|---|
2020.09.01 8:00-9:00 | 8 | No | N/A | 0 |
2020.09.01 9:00-10:00 | 15 | Yes | 5 | 5 × 0.9 = USD 4.5 |
2020.09.01 10:00-11:00 | 24 | Yes | 10 | 10 × 0.9 = USD 9 |
2020.09.01 11:00-12:00 | 20 | Yes | 10 | 10 × 0.9 = USD 9 |
The maximum elastic QPS is calculated by using the following formula: Maximum QPS of actual concurrent requests - Recommendation request QPS quota that you specified for your subscription AIRec instance. If the maximum QPS of actual concurrent requests exceeds the supported value, the system charges the elastic QPS based on the maximum elastic QPS. You can refer to the fourth row (2020.09.01 10:00-11:00) in the preceding table for an example.
Note: If the maximum QPS of actual concurrent requests exceeds the supported value, we recommend that you upgrade your AIRec instance.
Methods to modify a quota
To modify the recommendation request QPS quota of your subscription AIRec instance, log on to the AIRec console and click Upgrade Quota or Downgrade Quota on the basic information page of your instance.
Modifying the elastic QPS quota causes major changes. Therefore, you can modify this quota only every ten minutes.
Note: The elastic QPS used within an hour is charged based on the maximum elastic QPS in this hour. We recommend that you do not frequently modify your QPS quota.
Enable or disable elastic QPS
When you purchase an AIRec instance for the first time, you can enable elastic QPS. If you have purchased an AIRec instance, you can upgrade or downgrade your instance to modify the elastic QPS quota.