It may sound counterintuitive, but throttling connections is Server Management 101 type of stuff. So, unless you're hammering down on one IP address with the full weight of your connection, and that connection is big enough to fill your pipes, it's not going to be something you actually notice. Those many threads can be across multiple cores but no individual thread can be across multiple cores. Other connections would have their own thread.
That's one IP address sending one bit of data, like the HTTP requests. In BSD world, that's not allowed - due to an emphasis on rock-solid stability. If the throughput wasn't limited, it'd be possible for that thread to overwhelm the system and degrade the systems performance overall. I believe that's by design and has to do with the way that BSD schedules tasks.
0 kommentar(er)
