On several applications (most of them are web based) we see that the most of the slowest operations consist of (mostly) idle time. This, on one hand, makes the operation times longer and affect all statistics and on the other hand, as I understand, may suggest that the client itself is doing something locally.
I'm interested to hear your experience and best practices to better display these statistics and indicate problems.
Solved! Go to Solution.
The key value of this metric is to make the Network Time measurement more precise. As you know, in pre-12.3 releases we were calculating Network Time as a difference between the Page Load Time and the Server Time. "Calculated" is the key word here: AMD measures PLT and ST, but not the NT which is calculated later. In 12.3 release the AMD measures the PLT, the ST and the Idle Time (and the Other Time which are periods where parallel activities were occurring). So the NT calculation is more precise now.
There is a training module on the APMU that explains the differers and the value in more details: Measurements 10 Minutes | Last updated December 2014
Idle time in Oracle Database traffic can mean that your App Server takes some time to process the database queries. I saw this pattern a few times in some of my customers, the "client" from Oracle Database is usually the App Servers itself. Therefore, idle time in this case is the time taken for the "client" which is the app server in this case to process the response from the database (need validation from others on this).
Thus when high idle time in Oracle Database is coupled with high server time in the App Server, you can isolate the App Server to be the fault domain.