> What's the difference between my benchmark and a server that receives a lot > of requests ? > For you, this use case doesn't happen on production ? Or you maybe you have > a tip to avoid that ?
First, I don't use Gunicorn's default (sync) worker for receiving request from client directly. It should be used behind nginx or similar buffering reverse proxy. Second, I don't use Gunicorn's default worker for high load. Nginx and uWSGI via unix domain socket is much faster than gunicorn's sync worker. Last, to benchmark classical web stack on single machine, `wrk -c200` is too high. Concurrent connection and concurrent request is different at all. -- INADA Naoki <[email protected]>
