...
In this PROX which essentially measures the duration of a 1 nanosecond heartbeat, it is clear that some process is producing fairly frequent interrupts that cause delays in the 10-50 usec and 50-100usec range on core 6. Also, Core 11 has some interrupts in the 10-50usec range: it is important to check this when the VSPERF deployment is running (for example, the BSwLV OVS-Vanilla config has:
GUEST_CORE_BINDING = [('#EVAL(6+2*#VMINDEX)', '#EVAL(7+2*#VMINDEX)')] and
VSWITCH_VHOST_NET_AFFINITIZATION = False with VSWITCH_VHOST_CPU_MAP = [4,5,8,11]
We can get these statistics on a per-second basis, from a TCP socket - there is syntax to request these stats. (also -k command, Xavier will share scripts which are better).
...