What is this?

This knowledgebase contains questions and answers about PRTG Network Monitor and network monitoring in general.

Learn more

PRTG Network Monitor

Intuitive to Use. Easy to manage.
More than 500,000 users rely on Paessler PRTG every day. Find out how you can reduce cost, increase QoS and ease planning, as well.

Free Download

Top Tags


View all Tags

Unusual State - amount of measuring points to determine Unusual State

Votes:

0

When determining Unusual State, PRTG checks 'hourly average for this hour of weekday' or '24h average for this weekday'. However, how is this average calculated? How long a time-period is being used for calculating the average?

I'm looking here at traffic with daily, weekly and yearly cycles. So I need to know how long a time-period is used for calculating the averages, so I can determine how big the deviation of the average can be before PRTG should flag the traffic 'Unusual'...

average traffic unusual

Created on Mar 13, 2017 5:41:19 PM



Best Answer

Accepted Answer

Votes:

0

Hi Corné,

I actually had to ask our developers for the exact answer, it is 50 weeks. If you have less than 50 weeks monitoring data, then only the available weeks will be used, but the maximum is 50.

Best regards.

Created on Mar 15, 2017 11:33:02 AM by  Dariusz Gorka [Paessler Support]



11 Replies

Votes:

0

Hi Corné,

You can find the settings that affect the timespan for the unusual detection under "Setup > System Administration > Monitoring > Unusual Detection". You can also find more information about every unusual detection option on the following site under "Unusual Detection": https://www.paessler.com/manuals/prtg/monitoring_settings

Best regards.

Created on Mar 14, 2017 11:46:43 AM by  Dariusz Gorka [Paessler Support]



Votes:

0

Hi Dariusz,

When I look there, all I can change is the deviation from the average. I can't see anything to change the time-period used to calculate the average...

Kind regards,

Corné van den Bosch

Created on Mar 14, 2017 1:15:50 PM



Votes:

0

Hi Corné,

Just an example for the "24h average is <80% or >120% of weekday average" Option, the average will be calculated for the last 24 hours. For the option "Hourly average is <80% or >120% of hour-of-day average", then the average is calculated for the last hour. So you can't change the time, you only can change the percentage which will be considered as unusual.

Best regards.

Created on Mar 14, 2017 2:48:44 PM by  Dariusz Gorka [Paessler Support]



Votes:

0

Hi Dariusz,

Daily (24h): The average of the day before is compared to previous daily averages of the same weekday Hourly: The average of the previous hour is compared to previous hourly averages of the same weekday and the same hour-of-day

The way I read this: - Unusual is always about the previous day / hour; so, not the current values? - Daily average is compared to previous daily averages of the same weekday (very nice when monitoring data with daily cycles); but how many previous daily averages are averaged to get a value to compare to the average of the day before (very important when we're talking about monthly cycles in the values monitored!)? - Hourly average is compared to previous hourly averages of the same time at the same weekday (very nice when monitoring data with daily cycles); but how many previous averages are averaged to get a value to compare the average of the previous hour with (once again, very important when monitoring data with monthly cycles!)?

Kind regards,

Corné van den Bosch

Created on Mar 14, 2017 3:50:50 PM



Votes:

0

Hi Corné,

For example it is 8:00 AM, then the average (only the one of the hour) of 6-7 AM will be compared to the last hour (7-8 AM). The same is valid for the 24 hours. For example it is 8:00 AM again, then the average of the last 24 hours will be compared to the 24 hours before. So there is just one average during the time frames.

Best regards

Created on Mar 14, 2017 7:51:58 PM by  Dariusz Gorka [Paessler Support]



Votes:

0

Hi Dariusz,

What you're saying doesn't match what the description is saying.

According to you, the one-hour average is compared with the previous one-hour average, and the one-day average is compared with the previous one-day average. The description is saying the one-hour average is compared to previous one-hour averages *of the same weekday and the same hour-of-day*, and the one-day average is compared to previous one-day averages *of the same weekday*.

So which one is true? Because that's going to make a big difference when monitoring traffic with day- and week-cycles!

Kind regards,

Corné van den Bosch

Created on Mar 15, 2017 8:45:48 AM



Votes:

0

Hi Corné,

For the 24h unusual detection:
Imagine it is Monday, then the last 24 hour average will be compared to the 24 hour average of the Monday of the last weeks. So the last 24 hour average is maybe "11,2", last weeks Monday average is "12,5" and the week before it was "10,5". The two values ("12,5" and "10,5") will be compared to the last 24 hour average ("11,2"). If it is an increase of X% or decrease of X%, the sensor will switch to unusual.

For the hourly average:
It's actually the same. The hourly average of the last hour will be compared to the hourly average of the same hour and of the same weekday before. So Image it is Monday, 8:00 AM, the hourly average will be then compared to the hourly average of the (for example) last two Mondays, 8:00 AM.

Best regards.

Created on Mar 15, 2017 10:26:16 AM by  Dariusz Gorka [Paessler Support]



Votes:

0

Hi Darius,

Ah, now we're getting somewhere.

Question: How many previous averages are averaged to compare with?

"for example the last two" But how many exactly?

Kind regards,

Corné van den Bosch

Created on Mar 15, 2017 11:12:07 AM



Accepted Answer

Votes:

0

Hi Corné,

I actually had to ask our developers for the exact answer, it is 50 weeks. If you have less than 50 weeks monitoring data, then only the available weeks will be used, but the maximum is 50.

Best regards.

Created on Mar 15, 2017 11:33:02 AM by  Dariusz Gorka [Paessler Support]



Votes:

0

Hi Dariusz,

is there any way to change the value of 50 weeks? eg in a database?

Best regards

Created on Feb 18, 2020 3:32:38 PM



Votes:

0

Hi notino,

No, the value is hard coded and can't be changed. If there is less data available this number is lower but it can't be increased though.

Best,
Sebastian

Created on Feb 19, 2020 8:49:36 AM by  Sebastian Kniege [Paessler Support]




Disclaimer: The information in the Paessler Knowledge Base comes without warranty of any kind. Use at your own risk. Before applying any instructions please exercise proper system administrator housekeeping. You must make sure that a proper backup of all your data is available.