we have come across very strange thing the over all size and individual sizes of all AAS cube databases where reduced with out any changes happened. Did anyone saw this change dip and any cause?


it should have resolved now, Microsoft mentioned that was bug which hit most of the regions , we are on WEST US and MS released patch to make the size look accurate now


Yes! We've been having intermittent issues with this as well. Specifically what was happening was large spikes in memory usage. These spikes did not align with the timing of when our models were set to refresh but rather seems to have happened at random.

When models refreshed we would see an uptick in memory usage but it would fall after the model finished refreshing. That's normal. When we reached out to Microsoft their official response was "restart the service". We weren't thrilled with the response needless to say.

What we have done to combat this is use the Process: Clear command in-conjunction with our model refreshes. Since then we've only had a few of these incidents.

  • sirwinny, what tier you are using? and the intermittent spikes in memory is due to your end users activity if there are no scheduled refreshes, but there could be multiple other issues. And my issue described is different from what you are having. Since microsoft released 1500 support there are some changes happened at the way they calculate the size of cube. which i am not sure about. – JayM Sep 20 '19 at 15:57
  • @JayM We're using the S2 tier. At first glance it sounds like we're suffering from similar issues; I didn't elaborate very well. We not only had large spikes but also large dips. Your comment on end user activity I don't think applies here. Out models are small, 2gb is our largest model. We were seeing spikes and dips of 28GB+. We're also using compatibility 1400. – sirwinny Sep 23 '19 at 15:16
  • sure process clear will definitely help reducing the memory used by cube database in-memory, but it will cause data unavailability for users, but you can also check try checking the unwanted columns and hidden columns from aas and remove them, and if memory it self is a problem please vertipaq analyzer to check where is your memory and per table size by SELECT DISTINCT [CATALOG_NAME] as [Database], [CUBE_NAME], DIMENSION_CAPTION AS [Dimension], DIMENSION_CARDINALITY AS [Count] FROM $system.MDSchema_Dimensions ORDER BY CUBE_NAME you need to change the default database name. – JayM May 4 '20 at 16:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.