On Sat, Nov 23, 2013 at 07:26:25AM -0200, Fred Pedrisa wrote: > > OK. The last point could slightly help in reducing the number of calls to > kqueue and aggregate more events at once. But FreeBSD's kqueue is really > fast so that should not change much. You really need to be able to pin the > processes to certain CPUs, as well as the interrupts. Unfortunately I cannot > be of any help here :-( > > But do you believe the CPU pinning will really make all this difference ?
It completely depends on the CPU architecture. It's common to see a 2x improvement compared to the worst case which tends to be the default one. But for this you need a CPU with a fast enough shared cache and to ensure that you get *all* your NICs interrupts on (set of) a core(s) that share the same cache as the one(s) running haproxy. > I > know how to do it, using pthread, because I am used with it, just a few > lines of code are able to make it. It might be worth experimenting if you know how to do it. Regards, Willy