On Feb 27, 2015, at 6:36 AM, Chaigneau, Nicolas <nicolas.chaigneau@capgemini.com> wrote:
Trying to benchmark our setup, we noticed the following: - FreeRADIUS cannot handle more than a handful of requests per second. - only one, or two threads are used at the same time.
Some modules cannot be multi-threaded. The server implements mutexes to be sure that the modules are used only by one thread. If that module is used, *and* it takes a long time to do it’s work, then all threads will block on that module.
I do not understand what happens in post-proxy. Is there some limitation on requests parallel processing in this section ?
No. And the rest module doesn’t behave any differently in post-proxy. What other modules are you running in post-proxy? Arran and I have discussed instrumenting the server to track time spent in a module. That’s useful, but would destroy performance in a production environment. Still, for debugging it can be very helpful. Alan DeKok.