Don't understand result of aggregation in table visualization

Elastic Stack Kibana
1

Hi folks,
I'm using elastic stack 6.2.4.

I've build a dashboard to support the analysis of our loadtests. In the selected interval I am plotting 1st and last event timestamp, count of events + throughput. But I don't understand the results kibana is giving me, I expect much different results when I am doing the calculation in Excel.

I defined the first part of the aggregation like the following:
image

image
image.png339×761 16.3 KB

image
image.png334×788 16.9 KB

image
image.png344×797 17.4 KB

...
PLEASE klick on the images to see the full config. the forum is cutting of the buttom of the images in overview!

The result looks like the following:
image

I've selected following range in kibana: 2018-11-14 16:08:00.000 to 2018-11-14 16:40:37.312

I know I first had a mind error, I hoped the throughput will be calculated based on min and max timestamp, but on 2nd thought kibana i am quite sure that kibana will take the time window set up in the dashboard.

Lets discuss the first line:
My time window (selected in kibana is 32min and 37sec. That makes 1957 second. In this time we find 181053 events. But kibana tells me, that I have a throughput of 181053 per hour or 109 events per second. If I calculate the avg throughput per second I would calculate 181053 / 1957s = 92.55 events per second. That difference to 109 events per second is way too much to be an rounding issue.

What does kibana calculate or how can I get kibana to calculate it my way?
Btw: If my loadtest is running over the hour change, kibana is dividing the avg count per hour by 2 - the number of involved hours.

Can anyone please shed some light?
Thanks Andreas

2

@asp, it's not entirely clear what you would like to see exactly. can you post an example of the table (from excell e.g.) of what your result should look like.

also, from your screenshot, it's not clear what aggregation you are exactly running. from the screenshot, we can only see the metrics you are computing for a given bucket.

3

Thanks for the reply.

That's why I added that comment to the thread:

If you klick on the screenshots you can see, that I am using a count metric and time bucket set Date histogram to hour, minute or second. From that Calculation I show the average Bucket.

Also interesting and curious. If I query the same time interval again, I have slightly different results.
image

I would expect the following (excel sheet with GERMAN numeric notation. So you need to swap dot and comma in your mind):

kibana time interval 14.11.2018 16:08 14.11.2018 16:40
kibana interval time in sec 1.957
count count per hour count per min count per sec
01 - TBM2 HighPrio Gwy to Tux 2018-11-14 16:08:20.000 2018-11-14 16:40:12.000 181053 333.056,11 5.550,94 92,52
02 - NON-DTCO processing (tux queue empty) 2018-11-14 16:08:21.000 2018-11-14 16:40:14.000 102964 189.407,46 3.156,79 52,61
03 - DTCO processing (tap_dispatcher_msg empty) 2018-11-14 16:13:37.000 2018-11-14 16:38:28.000 79247 145.778,85 2.429,65 40,49

Formular: count / "kibana interval time in sec" = "count per sec"
"count per min" = "count per sec" * 60
"count per hour = "count per sec" * 3600

Original goal was to use min and max timestamp from the table for avg. calculation, but I think that is not possible with kibana.

4

@thomasneirynck
Any update here?

5

any news? additional questions?

6

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