Percentage of CPU Usage report AIX Vs. OEM

Experts -

I do have a discrepancy  on %CPU usage reports between AIX and OEM. For last 31 days, OEM is reporting %CPU around 40% and below, where AIX reporting around 80% consistently for last 2 months and came down couple of days ago.  How can I prove that AIX report is wrong?

or, how can I see the AIX history of CPU Usage for last 2 weeks?

The AIX report was generated from NMON utility.

Thanks,
Sve.
LVL 18
sventhanAsked:
Who is Participating?
 
EikromanConnect With a Mentor Commented:
Ok.
The OEM screenshot is showing data for the "Last 31 Days"
The NMON shows data between May 13 - to May 20 (7 days).  
So right here we have big difference in the reports granularity. OEM will not displays peaks in this mode. To have a more meaningful comparison:
1) Click on "CPU utilization" under the first graphic
2) On the right at the "View Data" select "Custom"
3) Select a range that is similar to the NMON report. You may experiment by selecting exact ranges as in the NMON or even smaller, like from May 16 to May 20, this way OEM report will be much more verbose.

Also,
It's not clear what exactly is being measured in NMON graphics. Is it 10%? Then why 11? Is it for all cores or single core?

Keep in mind that OEM graphic displays CPU usage in percents for ALL CORES. Meaning that if you have 4 cores, and one core is at 100%, the OEM will show 25% and if 2 cores are being used, the graphic will show 50%. I'm not familiar with NMON, and do not know, how you generate this graphic, but in Linux, for example, the TOP command displays percentage for the CPU Core. If compared to OEM, it will produce 100% and 200%  against 25% and 50% in OEM.

Let's continue with OEM report that I described earlier and see if there is a similar pattern for smaller ranges.

I personally doubt that there is a bug in either NMON or OEM, it looks more like different metrics, granularity or visualization method
0
 
EikromanConnect With a Mentor Commented:
Hello,

Although I have no real thought about the problem in question, it may help if you provide full samples of the reports.

Usually, such discrepancies are often associated with different measured metrics for example overall CPU (NMON) usage and CPU consumed by a component like Oracle DB processes (by OEM).  

Another possibility, is when reports are for different time period. "Real-Time" monitoring will catch every peak utilization, while daily report or monthly will display average statistics and so on.

But since I have no knowledge of what report is actually being used, I can only speculate. Give us some screenshots or something.
0
 
sventhanAuthor Commented:
Please find the screen shots attached.

Thanks for looking.
cpu-cmpre.docx
0
 
Franck PachotConnect With a Mentor Commented:
Hi,

The NMON report is not very clear. Is it a %CPU ? It scales from 0 to eleven !

But it seems that the report measures the 'Physical CPU'

OEM is measuring what the OS sees, and that's logical CPU.

It guess that the difference comes from that. 80% of 16 physical CPU is the same load as 40% of 32 Logical CPU. You should have the numbers in v$osstat to confirm.

Regards,
Franck.
0
 
sventhanAuthor Commented:
I've requested that this question be deleted for the following reason:

Need to edit the question
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.