Hi Ethan, > One more thing that occurred to me on the issue of lock contention. > Are you running with hyperthreading enabled? If so, you're going to have > quite a > few more threads hitting various locks, but not necessarily more CPU for them > to > take advantage of. I definitely would not expect our performance to scale > linearly > once we've got more threads than the number of real cores on the system. It > might be worth manually reducing the number of handlers and checking what kind > of numbers you get.
Yes, the processor is running with hyperthreading enabled, and total 32 cores after HT enabled. I have some data collected, please see table below. The input traffic load and pattern is same for all cases, and no any packets lost in either kernel or in dispatcher (miss_queue_overflow) since miss_handlers == 3, but we can see that the cpu usage has been 750% in 16 threads which are same number as physical cores. miss-handlers Total ovs-vswitchd cpu usage 2 340% 3 400% 4 439% 8 643% 12 726% 16 750% 32 840% Regards, CY. _______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
