1

I'm doing a CPU test using the stress-ng tool. But I want to understand what their output means. Can you explain, please? I added 2 screenshots.

For example:

AWS

DO

Thanks for your help.

mrbean007
  • 21
  • 2

1 Answers1

1

Each stressor will loop over a stress operation multiple times, the number of loops is the bogo-ops metric. "Bogus operations per second".

The "real time" measurement is how long the run took based on the wall clock time (that is, the time the stressor took to run).

The "usr time" measurement is the amount of CPU used running as a normal user process. Note with 4 CPUs this is about 4 x the wall clock time (since no sys time is being used).

Note that you have a space between the "-" and the "io" stressor option, so this is skipped. You need to run it using "--io 2" if you want to invoke that stressor or remove it if you don't. The io stressor was added to stress-ng to make it compatible with the older stress tool. If you want a good mix of I/O activity use instead the --iomix option or the --hdd option, e.g. stress-ng --cpu 4 --iomix 2 --timeout 120s --metrics.

The "sys time" measurement is the amount of CPU used in the kernel (aka system time).

The "bogo-ops/s (real time)" is a the bogo-ops rate: it is the bogo-ops count divided by the real time (run duration). This is the "apparent" throughput as if you measured the bogo-ops using a stop watch.

The "bogo-ops/s (usr+sys time)" is the bogo-ops rate divided by the user + system time. This is the real per CPU throughput taking into consideration all the CPUs used and all the time consumed by the stressor and kernel time.

Note that if you request a stress test to be run for 120 seconds stress-ng will try to ensure the stress test is terminated after 120 seconds but there are always small over-runs due to when the stress test checks to be terminated and/or has to do some clean-up after stopping.

Colin King
  • 121