Forum Discussion
How to avoid poor performance when the data base has millions of records
You should not use Direct Query if you can avoid it. Import mode is more performant. Here are some key things to consider:
- Use import mode
- Create a star schema data model
- Only bring in the data you need (remove any unneeded columns/rows)
- Avoid DateTime and Decimal columns (split into Date and Time, used Fixed Decimal)
- Leverage incremental refresh to minimize refresh times
Pat
Hi Anonymous
It's best to do most of the transformation on the source (create a view for that) and connect directly to the source using Direct Query. You can refer few of the posts below for more clarity.
https://www.youtube.com/watch?v=_zYvybVMk7k
https://learn.microsoft.com/en-us/power-bi/enterprise/service-premium-large-models
https://learn.microsoft.com/en-us/power-bi/guidance/power-bi-optimization
I hope this helps.
Hello,
thanks for your feedback!
I am little bit confused.
You said that i should use "Direct Query", but on one of the sites you shared (zeabra bi) says that using direct query can be very slow!
Should I or shouldn't I use direct query?
Something is escaping me!!
Best regards.
Hi Anonymous
Sorry for the confusion. You can fall back on the DQ approach if you need the most recent data, if you can wait for a few hours then Import mode would be best. I agree with the points shared by ppm1.