I did similar benchmark axum-hyper (was hyper below version 1.0) - using ab test utils, hyper was faster 2-3ms per request. I use hyper/tokio pair ever since, and never look back. Also - custom router is not that hard to add to hyper server, so You should not be confined to single endpoint server.
what if we test with real network call .If we are using localhost it doesn’t show the actual results !example aws api server with Asia region and benchmark from the europe or USA region.
I don't think we want the overhead of the network. We want pure code performance. In any case, the setup is the same for both APIs, so we have the same comparison baseline.
I did similar benchmark axum-hyper (was hyper below version 1.0) - using ab test utils, hyper was faster 2-3ms per request. I use hyper/tokio pair ever since, and never look back. Also - custom router is not that hard to add to hyper server, so You should not be confined to single endpoint server.
what if we test with real network call .If we are using localhost it doesn’t show the actual results !example aws api server with Asia region and benchmark from the europe or USA region.
I don't think we want the overhead of the network. We want pure code performance.
In any case, the setup is the same for both APIs, so we have the same comparison baseline.