Performance test tools in repo. about amqprs

[gftea]

@jacobmiller22 I setup a benchmark comparison between amqprs vs lapin, and it shows amqprs performance is much better than lapin in a basic_publish case.
if you are interesting, you can check https://github.com/gftea/amqprs/actions/runs/4142767627/jobs/7163827661

[michaelklishin]

You can use PerfTest for baseline comparison or develop a minimalistic implementation of PerfTest (the library part).

It's a non-trivial amount of effort but Team RabbitMQ did it for .NET for similar reasons.

[gftea]

@jacobmiller22 definitely a good proposal to go, performance engineering is never easy, but to have something better than nothing.

Thank you @michaelklishin to point out direction. Where can we access the baseline data?

[michaelklishin]

We don't really store baseline data, you would have to run PerfTest in your environment yourself. A basic 1 consumer, one producer, one quorum queue kind of workload would be enough to start.

[gftea]

I take a quick look of perftest, it seem by default using java client and mainly target for server performance test.

[michaelklishin]

Yes but you can also use it as a library and certainly can use it to establish a baseline for your client

[McPatate]

We use https://docs.rs/criterion/latest/criterion/ @ HuggingFace for our tokenizers library :)

[jacobmiller22]

@gftea Thanks for working on this. I was meaning to contribute, but just got caught up in some other personal stuff. Congrats on the better performance! If you have any suggestions on how to improve the test tools or any quality of life, I'd love to give it a stab.

[gftea]

I have added 3 different benchmarking methods, I will close the ticket now considering this is to first introduction of benchmarking.
For details, see https://github.com/gftea/amqprs/blob/main/benchmarks/README.md