True, and the longer answer is (and I will also invite all the engineers and architects to correct me if and where needed): you don’t need those because the Service VM is not involved in the scheduling of the RTVM, it is running directly on dedicated CPU cores and that configuration is static. Even the interrupt distribution is passed-through (using the –lapic_pt flag).

I would also add that there are many parameters that should be taken into account when you design a system using RTVMs, we have a number of recommendations and tips listed here [1] but it is not exhaustive and we always, of course, recommend that you do your own testing and analysis of your systems when it’s running.

[1] https://projectacrn.github.io/latest/tutorials/rtvm_performance_tips.html

Cheers,

Geoffroy