Forum Discussion
how to handle large data size
Hi,
I have a data size of 50 million records (only required columns considered) approx in sql server. what should be the best way (import or direct query) to build the report on this data size without any performance issue or slowness? Is adding a parameter good idea when connecting thorugh direct query?
Thanks,,,
Anonymous
I tend to avoid direct query if performance is important as unless you are filtering down to small data sets on well indexed sources it can be slow.
Take a look at aggregations as here you can get the best of both worlds by importing aggregations and direct query for individual records
https://docs.microsoft.com/en-us/power-bi/desktop-aggregations
https://www.sqlbi.com/tv/aggregations-in-power-bi/https://radacad.com/power-bi-aggregation-step-1-create-the-aggregated-table
3 Replies
- stretcharmMemorable Member
I tend to avoid direct query if performance is important as unless you are filtering down to small data sets on well indexed sources it can be slow.
Take a look at aggregations as here you can get the best of both worlds by importing aggregations and direct query for individual records
https://docs.microsoft.com/en-us/power-bi/desktop-aggregations
https://www.sqlbi.com/tv/aggregations-in-power-bi/https://radacad.com/power-bi-aggregation-step-1-create-the-aggregated-table
- kcantorCommunity Champion
Anonymous
It really depends on what type of data you have. PowerQuery is really good and compressing data. In addition if you research the query folding data, you will be able to determine the best place for your query to fold to maintain performance. Here are a couple of good resourses for you.
https://powerpivotpro.com/2015/12/compression-with-power-pivot/
https://powerpivotpro.com/2010/02/surprising-example-of-powerpivot-compression/
https://www.mssqltips.com/sqlservertip/3635/query-folding-in-power-query-to-improve-performance/
https://blog.pragmaticworks.com/power-bi-checking-query-folding-with-view-native-query
- AnonymousNot applicable
Hi there. I have a model with about 80 000 000 rows in the fact table and would never even consider the DirectQuery mode if I can use Import. The Import mode is THE BEST OUT THERE you can have. You cannot beat it using the DirectQuery or Dual mode because PQ compresses data with a factor of between 10 and 100x and the data is all stored in memory. YOU JUST CANNOT BEAT IT...
It's as easy as that.
Best
Darek