Forum Discussion
Lakehouse Table Rows Limit
Ok thanks for confirming. At the moment yes that is a limit as the directlake feature is (attempting to) paging all the rows into the vertipaq engine cache from the lakehouse table. I don't know whether it is exactly 1.5B rows as there could be a little variance. However if your Fact table is well above that, then yes it may need splitting if you want DirectLake.
You should be able to use a measure to Sum/Count across multiple tables and it should use directlake
You can use Profiler to actually see if a query is using DirectLake or falling back to DQ in your testing.
Learn how to analyze query processing for Direct Lake datasets - Power BI | Microsoft Learn
Hi gp10, the row limit of 1.5B rows is based on the DirectLake functionality - is this what you are referring to? The Lakehouse table itself can hold more, it's just that the DirectLake feature will not work with tables > 1.5B rows and will fall back to DirectQuery.
Thanks AndyDDC , yes this is what I'm referring to.
In this case, what options do I have if I want to use Direct Lake with tables more than 1.5b rows?
Is there a hard limit on the Lakehouse table rows?
Cheers.Ok thanks for confirming. At the moment yes that is a limit as the directlake feature is (attempting to) paging all the rows into the vertipaq engine cache from the lakehouse table. I don't know whether it is exactly 1.5B rows as there could be a little variance. However if your Fact table is well above that, then yes it may need splitting if you want DirectLake.
You should be able to use a measure to Sum/Count across multiple tables and it should use directlake
You can use Profiler to actually see if a query is using DirectLake or falling back to DQ in your testing.
Learn how to analyze query processing for Direct Lake datasets - Power BI | Microsoft Learn
Thanks for your response AndyDDC.
I will accept it as a solution, so far no other alternatives (other than upgrading the capacity) to use Direct Lake with such a table.In your opinion, how would the performance be if I use such measures?
And performance wise, when splitting the table, does it make sense to create few big tables that will try to approach the 1.5bn row limit, or create many smaller ones?
Thanks for your time.
