SLO differences with multidimensional analysis

[DamianG]

I'm working with SLOs to calculate % of requests with response time < 1s, defined as follows:

In the data explorer:

The SLO using the Dynatrace template is 95,6%.

If I perform this calculation using Multidimensional Analysis, the result is not the same:

• Number of requests with RT <1s: 5.16M

• Number of total requests: 5.18M

The SLO using manual calculations is: (5.16M/5.18M)x100=99.61%

Why this difference? I think this is related with the "partition" trasformation. How can I get the "real" value using SLOs?

Thank you.

[erh_inetum]

Hi Damián,

I think that the difference is because of the aggregation. On Data Explorer you are using "avg" and on MDA you're using "count". Try to use "count" in Data Explorer to check that the results are the same that in MDA changing the aggregation to "Count".

Regards,

Elena.

[DamianG]

Hi Elena,

It think this is not the case, because the "avg" aggregation in the Data Explorer is to get the number of requests with response time < 1s. I think they are counted with the aggregation "count" after the "splitby":

((builtin:service.response.time:filter(in("dt.entity.service",entitySelector("type(~"SERVICE~"),entityName.startswith(~"CE_~")"))):avg:partition(latency,value(good,lt(1000000))):splitBy():count:default(0))/(builtin:service.response.time:filter(in("dt.entity.service",entitySelector("type(~"SERVICE~"),entityName.startswith(~"CE_~")"))):avg:splitBy():count)*(100)):default(100,always)

In fact, if in the MDMA I use the "Response time" metric the value is the same as in my previous post:

Please note that my SLO is based in the Dynatrace SLO template for this cases:

Please correct me if I'm wrong.

Thank you!