Performance issues when using ADX option with promoted tags in Microsoft FinOps Hub

[gmallum]

Hi,
We are using Microsoft FinOps Hub and have enabled the ADX option for data storage. While this works very well in general, we are experiencing significant performance issues when using promoted tags in our reports.

Tags are essential for us to get meaningful insights, but when they are used in combination with ADX, the visuals in the reports take up to one minute to load or refresh. This makes the reports practically unusable.

We have tried scaling up ADX, which helps slightly, but the cost increase is substantial and still doesn't bring performance to an acceptable level.

Are there any known optimizations, workarounds, or planned improvements to address this issue?

[RolandKrummenacher]

Hi @gmallum. Are you running your own custom reports (with your custom KQL queries) on top of the ADX cluster?

[gmallum]

Hi @RolandKrummenacher,
No, only promoting tags in the editor in Power BI Desktop. Two tags are promoted. And using the Tags example page in Cost Summary report and Invoicing and chargeback report.

[RolandKrummenacher]

Hi @gmallum

In the Costs query, you can try setting Handle_mixed_case_tags and Trim_spaces_from_tags to FALSE:

[gmallum]

Thanks! I did play around with these a while ago but saw now that the mixed_case_tags was TRUE and it did help to set it to FALSE.

I have also created some reports on my own based on the semantic model from the 'Cost Summary'. I did a test and changed it to the 'Invoicing and chargeback' model, which has the same connection to ADX. And the performance was a lot better. Doesn't make any sense to me, guess I need to look into it.

[flanakin]

The connections are all dynamic based on the columns that are selected. I suspect you're not seeing issues because of the columns used. I'm definitely aware that the current promoted tags implementation has perf and scale challenges with larger datasets. We intend to replace it, but just haven't gotten around to that yet. We want a more complete allocation solution, which would handle tags better. I believe that's still in the backlog, but I'd have to go dig it up. Either way, feel free to create an issue for any feedback to help us prioritize things.