The minimum specifications for the system are completely dependent on the amount of traffic throughput required. Total throughput increases with the amount of CPU added, therefore at least t2.medium is recommended. Amazon provides a guideline Elastic Compute Unit (ECU) to give a comparable measure between different EC2 instances.
| Note | Typically, API Gateway performance is CPU bound, therefore the best use of system resources is to add extra ECU to an instance. |
A number of performance tests have been performed on different sized EC2 instances to estimate how many transactions per second API Gateway can process on a particular virtual hardware. The following example results are for a simple test that verifies HTTP authentication and forwards a 10K message to a backend service and returns to the client. The connection is over SSL with 2048-bit encryption. The results of this test are as follows:
| EC2 instance | Transactions per second |
|---|---|
|
m1.large (4 ECU, 2 Cores, 7.5GB RAM) |
240 |
|
c1.xlarge (20 ECU, 8 Cores, 7GB RAM) |
820 |
|
m2.4xlarge (26 ECU, 8 Cores, 68GB RAM) |
836 |