[RFE] Getting tuned to function closer to the older static way of setting system performance mode low cstate, so no added cpu cycles

Issue

Low latency shops especially in the FSI space try to minimize cpu cycles at all costs.

There's a push for customers to use tuned to get desired latency performance settings because it does a nice job of encapsulating those settings, however there is a concern about introducing added cpu cycles.

So for shops that don't need dynamic tuning and want to mimimize any added daemon cpu cycles, how can they utilize tuned without concern about the added daemon and threads that it introduces.

1. What's the overhead of having tuned daemon running?
2. What does tuned-adm latency-performance really change under the hood?
3. If we don't want to run tuned daemon, how can we do what we accomplish the same thing that tuned-adm latency-performance does?