Showing results for 
Show  only  | Search instead for 
Did you mean: 

Relative and Absolute Threshold


Hi All,

I have a question related to the automatic threshold baseline and also the relative and absolute thresholds.


The first question is how Dynatrace calculates the automatic baseline, is the learning time 7 days?


And what exactly are the relative and absolute thresholds? I can understand how they are calculated, but I couldn't understand exactly what these are. Can you clarify it for me? Thank you.


Best Regards,



Dynatrace Champion
Dynatrace Champion

The baseline is calculated over a 7 day rolling window. The calculation starts immediately and gets more refined over time until it can finally compare the current Monday to the previous Monday when it has a week of data. Low/High Traffic problems will not be raised until a week has passed since initial data collection for the application/service. Response Time and Failure problems will be raised after 20% of a week.

Relative and Absolute refer to thresholds set based on the deviation from the baseline, not the baseline itself.

An example of absolute would be 100 ms. An example of relative would be 50%.

Absolute means a concrete number. Relative means a percentage increase.

More details here:

Hope this helps,


Hi Dave,

So, focusing on Response Time, absolute theshold means that if the response time of baseline in the last 7 days exceded of 100ms it raises an alert, right?

And, for example, if we increase the absolute threshold from 100ms to 1000ms, an alert was raised if the response time of the baseline exceded of 1 sec in the last 7 days, right?



The baselining is not taking the average or median of the 7 days.

The baseline detection mechanism employs observation periods of 5 minutes and 15 minutes. This approach ensures that any detected baseline event must endure for at least 5 minutes, thereby preventing the generation of alerts for transient metric fluctuations.

On that is this calc right showcased over AD?



Threshold for failurerate and response time of service has configured AI based. To avoid over alerting, atleast 10 request should be available for 1 min and abnormal should be available for 5 min period( can able  to change delay)
Response Time:
Two conditions ( All req and slowest 10%) are available. It will create problem if either of the condition meets
All requests
Alert if the median response time of all requests degrades beyond both the absolute(100ms) and relative thresholds(50%):
Slowest 10%
Alert if the response time of the slowest 10% of requests degrades beyond both the absolute(1000ms) and relative thresholds(100%):
Failure Rate:
Alert if the percentage of failing service calls increases by both the absolute(0%) and relative thresholds(50%):
Absolute and Relative thresholds:
Absolute threshold means the threshold we set manually. Relative threshold is the deviation % from auto generated baseline's reference value. Auto generated baseline has been detected by past 7 days data and baseline will be different for each request or htm and each service.
Example 1:  Resptime: 100 ms of absolute threshold and 50% relative. auto baseline detected is 60ms. actual resptime is 95ms. It will not generate alert. Because value should cross both absolute and relative.
absolute 100 ms , Relative = 60ms (auto baseline) + 30 ms (50% of auto baseline) = 90 ms. So if actual respTime is 95 ms. It will not trigger alert. Because relative breached but absolute value is not breached.
Example 2: FailureRate: 0% of absolute threshold and 50% relative. auto baseline detected is 0.15%. actual failure% is 0.45%. It will generate alert. Because value has crossed the both absolute and relative.
absolute 0% , Relative = 0.15% (auto baseline) + 0.075 (50% of auto baseline) = 0.23%. So if actual Failure rate is 0.45%. It will trigger alert. Because relative and absolute value is breached.

Featured Posts