Per machine, that is 4,250k rps. For any http service we write in Go, that lines up with our general sizing estimates before we start work on a generic new serv

Per machine, that is 4,250k rps. For any http service we write in Go, that lines up with our general sizing estimates before we start work on a generic new service. After we start to hit the service with load, we start profiling and go from there. Sometimes it gets faster and sometimes it gets slower, but it is usually in the right ball park.