Alerts based on historical values

Alerting based on historical values is robust against seasonality changes, e.g. low usage on weekend or high usage during high peak business hours, because the current time range is always compared with the same time range of the same week day of the last n weeks.

You can specify for each self-defined group if you want to get notified when the average response time of the current time range is x percentage higher than the same time range in the past.

The following fields define the behavior:

  • Number of weeks: How many weeks (between 1 and 9) should be read from past for that time range.
  • Alert Threshold (in %): Define when an alert is raised in % of the historical avg. response time.
    Default is 200% ( = 2 times higher).
  • Time Range [min]: Defines how often the alert should be calculated. Default is 10 minutes
  • Minimum Number of Executions: Define how many executions must be available in the interval. Default is 10
  • No Auto Confirmation on Green Alert: Define if an alert is automatically confirmed after green occurrence.
  • Notification Variant: In case someone should be notified you can choose a variant.
  • Outbound Connection Variant: In case something should be triggered automatically you can choose the variant.

The following picture illustrates under which conditions an alert is raised with an Alert Threshold of 200%.

You can define additionally under Group Configuration not only dedicated systems but also LMDB Attributes:

  • IT Admin Role: e.g Production System
  • Priority: e.g. Very High, High
  • Lifecycle Status: e.g. Active, Installed
  • Additional Attribute Key: Your own defined LMDB attribute
  • Additional Attribute Value: The value you have entered for that system under the LMDB attribute

All Attributes are AND combined, different values per attribute are OR combined.

Important!

  • If systems are entered only these systems are taken into count for alert calculation.
  • If LMDB Attributes are entered only the systems where all the entered combinations matches are taken into count for alert calculation.
  • If neither system IDs nor LMDB Attributes are entered the alert calculation takes all systems for alert calculation.

 

In the picture above are all Systems selected where

IT Admin role is Production System AND

Priority is Very High OR High AND

Lifecycle Status is Active OR Installed AND

Additional Attribute Location is set

(Attribute Value is not not evaluated because it is empty)

Some hints:

  • This alert is good for recurring executions. It takes always the last completed time range. If the time range is set to 10 minutes it takes values from min 00 to 10. Next time range would be from minute 11 to 20.
  • Do not set the time range to small then you might get to many deviations and don't set too big because then the delay until alert is calculated might be too long. A good starting point is between 10 and 60 minutes.
  • Mixing different request types usually makes no sense, because they have different response time ranges. Front-end requests are usually around some seconds whereas back-end requests are around hundreds of milli-seconds.
  • Limit the amount of possible alerts by entering values as precise as possible. Wild cards are allowed but to generic values could create too many alerts.