Skip to main content

I have realized a strong effect on graph scales:
On the web interface periodically measured mean values (e.g. INTERFACE_TRAFFIC / CPU Load ) are displayed with correct peaks (Y-values) only when specifying an observation time period of less than 6 ¾ hours (3645 minutes).

Displayed maximum peaks shrink seriously (due to coarsening) the more the observed time period grows:
1. 6 ½ hours:    668 mbit = 100 %
2. 6 ¾ h hours: 569 mbit =   85 %
3. 24 hours:      366 mbit =   56 %
3. 1 week:           92 mbit =   14 %
4. 1 month:         50 mbit =     7 %
5. 1 year:            36 mbit =     5 %

When checking for spare operating capacity, this can be quite deceptive.

Hello,

Would it be possible to provide us a "How to test" or some screenshots so that we can reproduce the behavior that you describe to us internally? this would help us to understand this case clearly and find a possible solution

 

Thank you very much for your feedback and thank you in advance for your future response.

 

Best regards Baptiste

Sure. We use Centreon to visualize data transmission rate at the ports and work load of our Cisco network switches over time. Customizing is provided by a supplied service. Here are some graphs generated from the same collected data but with variant time frames. Sampling rate is 2 minutes. Complained effect can be seen on outliers: Peaks in number series shrink the more observed time intervall is enlarged. Measured maximum y-values are only visualized when displayed time slot smaller than 6 ½ hours. Have a special look at the vertical scale and value display (yellow marks):

  • Centreon observation intervall 1 day:

     

  • Centreon observation intervall 1 week:

     

  • Centreon observation intervall 1 month:

     

Here comes a counter visualization example (what I actually want to get) from another vendor, which does not shrink any peaks when extending observation time frame onto a whole year - a peak of 40 MBit/s stays always the same - a so called “max-based graph”:

  • Unify observation intervall 1 month:

     

  • Unify observation intervall 1 year:

Thank you very much for your interest - I’ve provided these screenshots for discussions already some time ago.

Reply


Terms & ConditionsCookie settings