Benchmarking for Continuous Optimization

I have it enabled in a couple of places, actually, quite loaded once recently. It was considered as low risk and threshold and sample rate. We need to run benchmarks and understand what's the price here. How would we do it in some system? For example, we have some system, OTP. And we want to measure it. Well, yeah, but this is specific benchmarking. But definitely, latency is the main point here. Yeah. The main danger in this benchmark is to bump into some different limit. Like disk IO or anything else and measure it instead of real overhead from auto explain. You don't want to reach the edge in terms of CPU and so on.