Azure Analysis Server becomes unresponsive if tables are getting processed

Question

We have multiple Power BI reports hosted on a server(S9V2) which has multiple databases. We have been noticing that whenever entities are getting processed, the reports are becoming very slow and even the processing of entities take more time than usual. This issue mostly occurs at a time when the user base accessing the reports are more (around 3-4K) The QPU's of the server at this time hit a peak and then it plateaus there for a long time. Is there any relation of processing a table when there are many users accessing the reports that can cause this? Or maybe a scenario where the server is not freeing up the resources (all allocated to the processing) for normal measures to run when users are opening reports. Any feedback is appreciated.

Answer 1

An S9V2 is a pretty powerful server to start with.
But yes, if you have too heavy a processing load, it can definitely interfere with concurrent query processing.
I find this is most prevalent when the processing involves a lot of CPU work.
A good example is if you have a large number of calculated columns. I try to move those back into the DW that I'm loading from, or see if I can move them to measures, so they are resolved at query time, and not as heavy CPU at processing time.
The other option is to break the load across multiple AS servers where that makes sense. As you're already at the top SKU, it's really the only option for scaling.

Answer 2

Hello Sachin Kumar

• One of the suggestions that we could think of is consider scheduling the table processing during off-peak hours when fewer users are accessing the reports.
• You may also consider scaling up the server to handle the increased load?