>Busy% = (Sum all Service units in the last 60 seconds) / (Num of
Processors * Cpu Speed per second * 60)
SU's are a software measurement.
And, as such, the amount accumulated per workload is different depending on
many things, such as dispatching priority, workload activity, swapability,
paging, phases of the moon, and the day of the week, etc. (8-{]}
All resource usage is not collected: there are only so many measurement points
in the system.
The difference is called uncaptured time.
>This formula seems rigth to me,
It's not, since the uncaptured time is missing.
>but Busy% is not even close of what RMF shows.
RMF is based on the hardware clock, and does not collect information anymore
granular than processor busy.
You should be seeing your equation generating between 75-85% of the RMF
reported number, as you call it.
(All the figures are from RMF data collection).
The variance depends on paging/swapping, and OS release levels, and other
factors, but those are the most prevelant.
I gave a more detailed explanation, in May 2009, under the title:
Re: RMF Monitor III view PU/PROCU question
If you're interested, hit the archives.
-
Too busy driving to stop for gas!
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
