Generic http vs async-http in ruby 3.x

[kinnalru]

After porting IO-bound (lot of aws-s3 uploads) project to falcon without any code modification it shows outstanding concurrency. S3 uploads take most of the request processing time.

After that I decided to change default http request to async-aws-ruby gem which uses Async::Http::Endpoint and Async::Http::Request. And there is no difference between official aws::s3(generic http) and patched async::aws::sdk.

Is there because ruby 3.x has better async support and there is no more need to use async-http in general case http scenario?

[kinnalru]

@ioquatix hi :)

[ioquatix]

Hello. Actually, I don't have a good answer to your question. However, the main advantage of correct usage of Async::HTTP is (1) HTTP/2 multiplexing, and (2) persistent connections amortise connect cost. However it must be implemented correctly and I don't know if it's true for the async aws implementation.

To see what is going on, you can enable debug logging. CONSOLE_LEVEL=debug falcon serve etc. It may give you some insight as to the internal behaviour. It's important to remember, web latency is significantly higher than processor instruction overhead. So even if one library is more or less efficient, it might not be visible.