Does AWS API Gateway HTTP know to work traffic spikes?
0
There is a known issue for AWS Classic and Applications LoadBalancers with traffic spikes - their autoscale is relatively slow.
Does AWS API HTTP Gateway suffer from the same issue?
Thanks, Vitaly
API Gateway has a default limit of 10,000 requests per second, with a burst of 5,000 requests. Usually, the bottleneck is not API Gateway, but the downstream integration it makes calls to.
Thanks!
"API Gateway has a default limit of 10,000 requests per second, with a burst of 5,000 requests." - yes, I saw that. But I guess under the hood API Gateway has some autoscaling mechanism, even though I couldn't see its details.
API Gateway is a multi tenant service. It is capable of handling much higher traffic than that, but it limits each customer to their specific limit. You do not need to worry about scaling API GW.
Regarding the backend, if you use HTTP API, and you use CloudMap as the integration, and all the instances are registered in CloudMap, I think you can omit the load balancer.
AWS OFFICIALUpdated a year ago
Thanks! "API Gateway has a default limit of 10,000 requests per second, with a burst of 5,000 requests." - yes, I saw that. But I guess under the hood API Gateway has some autoscaling mechanism, even though I couldn't see its details.
Regarding downstream integration - currently, I have API HTTP Gateway => ALB => ECS service. I'm thinking to use App Connect, i.e. exclude ALB. But here https://repost.aws/questions/QUfhuXtV8KRceDuC5fH_oMwA/is-running-fargate-and-api-gateway-without-a-load-balancer-in-the-production-environment-a-bad-idea I see two answers that it's a bad idea.
API Gateway is a multi tenant service. It is capable of handling much higher traffic than that, but it limits each customer to their specific limit. You do not need to worry about scaling API GW.
Regarding the backend, if you use HTTP API, and you use CloudMap as the integration, and all the instances are registered in CloudMap, I think you can omit the load balancer.