Dedicated instances allocate all cloud resources exclusively to a single tenant, making them ideal for production environments.
API Gateway offers eight instance types with the following performance metrics.
|
Instance type |
Maximum inbound requests per second (RPS) |
Maximum inbound connections |
Maximum inbound new connections per second (CPS) |
Maximum outbound connection pool size |
Maximum inbound public network bandwidth (Mbps) |
Maximum outbound public network bandwidth (Mbps) |
Cache |
SLA |
|
api.s1.small |
2,500 |
50,000 |
5,000 |
1,200 |
5120 MB |
100 Mbps |
1 GiB |
99.95% |
|
api.s1.medium |
5,000 |
100,000 |
5,000 |
2,400 |
5120 MB |
100 M |
2 GiB |
99.95% |
|
api.s2.large |
10,000 |
200,000 |
5,000 |
4,800 |
5120 MB |
200 M |
4 GiB |
99.99% |
|
api.s2.large.x2 |
20,000 |
400,000 |
10,000 |
9,600 |
5120 MB |
200 M |
8 GiB |
99.99% |
|
api.s2.large.x3 |
30,000 |
600,000 |
10,000 |
14,400 |
5120 M |
400 M |
12 GiB |
99.99% |
|
api.s2.large.x4 |
40,000 |
800,000 |
20,000 |
19,200 |
5120 MB |
400 |
16 GiB |
99.99% |
|
api.s2.large.x5 |
50,000 |
1,000,000 |
20,000 |
24,000 |
5120 MB |
600 M |
20 GiB |
99.99% |
|
api.s2.large.x6 |
60,000 |
1,000,000 |
20,000 |
28,800 |
5 GB |
600 M |
24 GiB |
99.99% |
-
HTTP transmits data serially over persistent connections—a sender must wait for acknowledgement before sending the next request. Use this to estimate whether the outbound connection pool size is sufficient.
-
The outbound connection pool connects API Gateway to your backend services. For example, an api.s1.small instance has a maximum pool size of 1,200. If each backend request takes 1 second, the instance supports a maximum outbound RPS of 1,200. When RPS exceeds this limit, new requests queue for a connection. If no connection is available within 500 ms, a D504CO error is returned to the client.
-
Additional constraints are documented in Limits.
Usage recommendations for dedicated instances
1. How do I select a subscription instance type?
API Gateway rates instance types by maximum RPS. For a given workload, QPS is typically higher than RPS, so you can use your QPS estimate to choose an instance type.
2. How do I select an instance for scenarios with traffic spikes, such as promotional events?
Two approaches handle traffic spikes. First, Upgrade/Downgrade an instance as needed. Second, combine a subscription instance with a pay-as-you-go instance using API group migration. For example, if your daily average is 2,000 QPS and you expect 4,000 QPS during a 24-hour event:
-
Purchase a yearly api.s1.small subscription instance for daily use.
-
Before the spike, purchase an api.s1.medium pay-as-you-go instance. Three hours before the event, switch the API group to this instance in the API Gateway console and verify access. After the event, switch back to the api.s1.small subscription instance, verify access, then release the pay-as-you-go instance. This minimizes additional cost.