Brett CannonAnyone know of a blog post comparing the performance of the various #Python async HTTP clients (e.g. aiohttp vs httpx vs ...)?
Gramster@brettcannon we have some basic GET benchmarks here: https://github.com/mikeharder/http-get-perf. I am adding httpx (which seems slower than aiohttp) and blacksheep (which seems faster).
Michael Manfre@geekraver @brettcannon what value is to be gained from running performance tests where they are not run with similar test configurations? Why do some have their threading/concurrency/connections set to different values?
@mike @brettcannon each system may have optimal settings to maximize rps. Not saying that those have been found in every case, but I certainly wouldn’t expect them to be the same across the board.
@geekraver @mike @brettcannon I agree that they would be different, but even if the optimal numbers were found the metrics would not be meaningful without explaining why those specific numbers were used. With concurrency being a runtime setting, you could run all the packages with 32, 64, 128. That would convey a lot more meaning without needing to find and/or explain an optimal value.