Rust in front of Django buys 311,000 requests a second on a static JSON endpoint

Django-Bolt's published figure is 311,000 requests per second ^[1]. Divide one second by that and each request gets about 3.2 microseconds of server time ^[8]. The endpoint being measured does not check a user's identity, query a database or run application logic, so none of that time is in the number ^[9].

Django-Bolt does not ask a team to give up Django. The existing application and its Python logic stay in place while more of the HTTP handling moves into Rust ^[2]. Once the Rust server hands the request over, Django still has the endpoint's real work to do, and the response may depend on a database query, application code or another service answering first ^[11].

The simple endpoint measures one thing well: how much overhead the framework itself adds ^[9]. It cannot tell a team how much faster an endpoint becomes when most of its time is already being spent inside Django or waiting for data ^[5].

So the question in front of a team with a slow API is where that endpoint's time currently goes. Profile the endpoint people actually depend on and split its response time into HTTP serving and everything else. If serving is a thin slice of that total, the improvement available is that slice, whatever the empty endpoint reports. The trade-off is real: this test takes longer to wire up than pointing a load generator at a static route, and it comes back with a smaller, less quotable number.

devops.com wrote that a lower number from that kind of test "may be far more useful than an impressive result from an empty endpoint, because it tells you whether Django-Bolt is speeding up a part of the application that was actually holding it back" ^[7].