All,
Here's a curiosity question, which I haven't found a satisfactory answer
for in the help panels or the manual.
On z/OS 1.10 in RMF III, looking at the PU/PROCU panel, and adding up
the "Time on CP% : Total" column, which is supposed to contain the
percent of the CPU used by the task during the sampling period, yields a
number over 100%. I got 157.6% in this case.
Granted, the LPAR has two logical CPUs, the CEC has three physical. So
then, why wasn't the number 200%? The LPAR wasn't anywhere near it's
cap. The CEC wasn't anywhere near 100% busy (thus, no short engine
effect could be causing this). The LPAR was getting about 58% of all
available physical CPU for the CEC, logical CPs getting about 89%
dispatched.
Any way that I've sliced it, I don't get an even 100% or multiple of
100, in looking at this. And the numbers are much too far off to be
explained by loss of precision.
I've gone so far as to select a timeframe where the CEC was at 100%
physical busy, and added the PU/PROCU TOTAL columns from all LPARs, and
still only came up with 249% (versus the 300% I'd expected for 3
physical CPUs).
Can someone please explain why adding the total column isn't yielding a
multiple of 100?
For your info, here is what the HELP panel says about that column:
RMF Monitor III Processor Usage - Field Help
COMMAND ===>
Time on CP % Total is the percentage of CPU time spent on general
purpose
CPs as sum of TCB time, global and local SRB time, and preemptible or
client SRB time consumed on behalf of this address space.
Thanks,
Gary Diehl
Systems Administration
"Water seeks it's own level" - Aristotle
----------------------------------------------------------------------
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
