Different %ProcessorTime metrics in Amazon and System

[AlbertoOrdoñez_García]

Hi,

i notice that %ProcessorTime graph is different in Amazon and System counters for the same machine. The pattern is more or less the same, but in System counter shows around 20% more of Processing Time. (see attached images, AmazonCounter,AmazonConsole and SystemCounter) Monitor Desktop is installed in the monitored machine and Monitor Server is in other machine.

Any idea on how to resolve this?

Thank you.

Alberto.

[AndrewSakowicz · Accepted solution]

There are two factors:

1. "Monitor Desktop is installed in the monitored machine"

This configuration adds expected "extra" cpu. Here is why.  When "Monitor Desktop" collects, the collector process (node.exe) will acquire temporarily one cpu.  If machine has 4 cores, this would amount to extra 25% to your normal cpu usage.  AWS resolution is 5 min and that spike is either missed or  "averaged out".  

2. collection interval 5 min AWS vs. 1 min System

5 min interval will not show the spikes visible in 1 min collection, e.g. 97CPU%

[AlbertoOrdoñez_García]

Thanks Andrew for your response. I assume that Monitor Desktop adds an extra charge to the system but, should AWS Counter and console detect this extra charge and show more or less a similar graph, P95 or Avg Statistics using the same collection interval? 

So, the main reason for this differences is a different collection interval. Is it correct?

Thank you again.

Alberto.