Why do CPU utilization metrics differ between "Logical Partition" and "CPU Summary" attribute groups in the ITM warehouse, for the AIX Premium agent?
For example, in the same monitoring cycle, "Idle CPU Pct" may be varying around 85% for the "CPU Summary" attribute group, while it is varying around 70% for the "Logical Partition" attribute group. The other CPU metrics (IO Wait, User and Kernel) are showing similar differences between both attribute groups.
When historical data collection is enabled for several attribute groups at the same time, there's a chance that each data collection puts a certain load on the system.
For example, suppose there are 5 attribute groups which are collecting historical data:
As "CPU Summary" is the first attribute group in the cycle, and "Logical Partition" the last attribute group in the cycle, this may explain why the values for "Logical Partition" always show a higher CPU load, compared to "CPU Summary".
After disabling historical data collection, the differences would persist, this may be IV18372.