The Priority Queueing feature deprioritizes the packets of an identified elephant/heavy flow when the CPU utilization of a individual Firewall Worker Instance reaches 100%. My question is for how long must the CPU utilization of that Firewall Worker Instance be at 100% before Priority Queueing kicks in? This doesn't appear to be documented; all the official documentation and SKs just say that Priority Queueing becomes active when the core reaches 100%.
Poking around in the Firewall Worker Instance kernel variables, there is an undocumented variable that may be related:
fwmultik_prio_grace_thld = 300
I'm assuming the 300 indicates 300 milliseconds, so is the answer 0.3 seconds? That seems like a reasonable value.
Gateway Performance Optimization R81.20 Course
now available at maxpowerfirewalls.com