Visualize and compare with normal daily trend to identify abnormal behavior


I want to monitor transaction response for one of my systems and compare it with the normal daily trend. Depending on the different time of the day, this response time can vary. I want to identify abnormal values of the response time. One possible way that I can think of is, to compare the current response time with the average response time for last five days for the same time.
eg: if current time is 08:05 AM, need to get the average response time at 08:05 AM for last 5 days and compare it with current response time and check the deviation. If deviation is within 10% consider it's healthy and acceptable.

I'm not sure how to implement this kind of behavior in Kibana and Elasticsearch. Seems like moving average is also not a solution for this. Really appreciate on any possible way to implement this sort of functionality or any other possible way to get the same outcome.


If you want to compare daily trends, these are your options:

  • Subscribe to paid machine learning features which will learn the trends of your data
  • For basic week-to-week comparison you can use TSVB to draw two lines (one with an offset)
  • For more advanced week-to-week comparison you use Timelion, which can be used in a dashboard or in Canvas
  • Canvas also allows you to build multiple queries and compare them in a way you choose

Thank you for your reply. I'll try these options

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.