on ‎2020 Sep 23 8:32 AM
Hi,
We are using the CPU workload analysis a lot to monitor peak periods.
The CPU part is rather clear but the waiting threads stats are not.
First of all, the number scale is rather confusing and we wonder if this work as it should. We look for instance at a window of 3 hours (eg 9-12 AM). In this example, the bars don't seem to reach 1K.
But one I move the time frame a bit (eg 9:15 - 12:15), the number scale goes a lot higher. The strange things that the bars for 10 and 11 AM (where the data shouldn't have changed) doesn't change according to the scale.
If I have eg 850 waiting threads at 10 AM, that should remain the same whatever the scale is. But the bar size doesn't resize, despite that the scale is changed to larger numbers.
Secondly, what is the relationship between the waiting threads and CPU usage? Since both are shown on one screen, a logic conclusion would be that the higher the number of waiting threads, the higher the CPU. That doesn't seem to be case, if one takes the example from above with the high number scale. It says that there is X k of waiting threads but the CPU usage is rather low.
We also done a test with replacing a query with a more performant one and it doesn't show in the waiting threads (and this the amount is lower) anymore, but there isn't a change CPU wise.
So what is the exact meaning of the waiting threads, should they be as low as possible an what is the relationship with the CPU usage?
Cheers,
Eddy
Request clarification before answering.
| User | Count |
|---|---|
| 13 | |
| 8 | |
| 7 | |
| 5 | |
| 4 | |
| 3 | |
| 2 | |
| 2 | |
| 2 | |
| 2 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.