Adding time derivatives of some metrics about prmon HOT 6 CLOSED

Hi Andrea,

If I understand your request correctly the core functionality should be there already, i.e.:

prmon_plot.py --input ... --xvar wtime --yvar rx_bytes

would create the usual time evolution of rx_bytes like:

while

prmon_plot.py --input ... --xvar wtime --yvar rx_bytes --diff

would compute and plot delta(yvar)/delta(xvar) as:

I think we can improve the labeling etc. but is this what you're after?

Best,
Serhan

from prmon.

Exactly! The only degree of freedom is the delta_t used to estimate the derivative. It should be a multiple of the sampling interval; some degree of smoothing may be desirable (maybe a sensible default and a configurable parameter? Up to you!).
Cheers,
Andrea

from prmon.

I was thinking again, and maybe it's not just a matter of plots, but we should consider IO rates as "real" metrics, which means having their values written already in prmon.txt, such that prmon.json will contain max and avg of them, too... Clearly it's more work because they should be calculated in prmon itself.
Andrea

from prmon.

Hi @sciaba - I am not so sure about that. From the accumulating values it's easy to calculate the derivative and we could easily adopt a few strategies to do that (various sliding window shapes and widths). If you take the derivative inside prmon and then only give rates it's not so easy to go backwards.

Technically it's not that hard to do, but I what we do now is more useful and reflects the underlying metric better, so it's easier to relate to other measurements.

from prmon.

You convinced me!

from prmon.

The plotting aspect of this issue is addressed in #66. I think the consensus is not to make any changes to the core code for this. Therefore, I'm closing this issue. Please feel free to re-open if need be.

from prmon.