This topic describes the billable items of Cloud-native API Gateway instances during automatic scale-outs. This helps you understand and calculate the fees that may be incurred when you use the auto scaling feature of Cloud-native API Gateway.
Capacity water levels
Cloud-native API Gateway supports auto scaling. You can configure a period for auto scaling to better cope with traffic surges. Scale-outs are automatically performed by unit. The following items describe the capacity of a unit:
Secure thresholds: On and beyond secure thresholds, your gateway instance can ensure high throughput and low latency even when traffic is doubled.
Warning thresholds: When your instance hits warning thresholds, it endures increased latency and is subject to stability risks in cases of traffic surges.
Scale-out unit | 1 unit | |
Client connections | Secure threshold | 12,000 |
Warning threshold | 24,000 | |
New HTTPS connections per second | Secure threshold | 400 |
Warning threshold | 800 | |
CPU utilization | Secure threshold | 30% |
Warning threshold | 60% | |
Memory usage | Secure threshold | 75% |
Warning threshold | 75% | |
QPS reference
The queries per second (QPS) throughput of a gateway instance is affected by various factors, such as the response size, HTTPS enabling, and GZIP compression. The following table lists the worst-case QPS values of different instance specifications based on a 30% CPU utilization:
New HTTPS connections consume large amounts of CPU resources. In business scenarios in which a large number of HTTPS connections need to be established at the same time, you can estimate your instance capacity requirements based on the data of short-lived connections in the following table:
Scale-out specification | 1 unit | |||
Connection type | Response size (KB) | HTTPS | GZIP | QPS reference at a CPU utilization of 30% |
Short-lived connection | 1 | No | No | 1,700 |
Yes | No | 500 | ||
Long-lived connection | 1 | No | No | 2,200 |
Yes | No | 2,000 | ||
Yes | Yes | 1,700 | ||
10 | No | No | 1,800 | |
Yes | No | 1,700 | ||
Yes | Yes | 1,000 | ||
Scale-outs of different instance specifications
Specification | Existing units | Scale-out step size | Scale-out range |
apigw.dev.x1 | 1 | N/A (This specification does not support auto scale-out.) | |
apigw.small.x1 | 2 | 1 unit | 3-4 units |
apigw.small.x2 | 4 | 1 unit | 5-8 units |
apigw.small.x4 | 8 | 1 unit | 9-16 units |
apigw.medium.x1 | 16 | 4 units | 20-32 units |
apigw.medium.x2 | 32 | 4 units | 36-64 units |
apigw.medium.x3 | 48 | 4 units | 52-96 units |
apigw.large.x1 | 64 | 4 units | 68-128 units |
apigw.large.x2 | 128 | 4 units | 132-256 units |
apigw.large.x3 | 192 | 4 units | 196-384 units |
apigw.large.x4 | 256 | 4 units | 260-512 units |
Billing description
For the elastic nodes that are automatically scaled out, the system separately charges fees on a pay-as-you-go basis. In the billing details, the fees generated for the elastic nodes are separately listed.
Region
Scale-out unit
Pay-as-you-go unit price (USD/hour)
Chinese mainland
Unit
0.146
China (Hong Kong) and Japan (Tokyo)
0.218
Singapore, Indonesia (Jarkata), and Germany (Frankfurt)
0.202
US (Virginia) and US (Silicon Valley)
0.174
Scale-out fee = Number of scaled-out units x Unit price x Duration in hours. For example, if your instance is scaled out by 2 units between 8:00 and 10:00 in a Chinese mainland region, the scale-out fee is 2 x 0.146 x 2 = USD 0.584.