Closed kmmelcher closed 4 months ago
it depends on several factors, the kernel being used, the scheduler being used, the scheduling HZ rate (250, 300 or 1000 Hz), the processor speed and number of processors. Should be within 60 seconds, but it may vary. Also, it's hard to get an exact load when processors have frequency scaling and turbo speeds since this changes the amount of work done per scheduling quanta so it may have some jitter. The code uses a first degree feedback loop, so it may suffer from some oscillation in the short term until it reaches a steady state.
I suggest using the stress-ng --vmstat 1 option to see how the load is varying over time.
Thanks @ColinIanKing!
I'm doing some experiments with stress-ng and I'm trying to understand its behaviour.
I'm running this command:
This will cause the stress-ng tool to run for 15min using 70% of one CPU right? But as far as I know the 70% cpu usage won't be achieved automatically since stress tests needs to start up to get to that desired cpu usage, right? Any idea how long does it take to get to that desired cpu? Are we talking about 1-2 minutes or 5-10 seconds?