Fabric Architecture with imports from Snowflake

[Sarien]

Hello,

I would like to understand how I can leverage the new Fabric capabilities in our solution.

We are currently using:

1. Snowflake as a data warehousing solution

2. Import into Power BI using a Power BI Dataflow

3. Then it moves into a Power BI dataset

4. Is visualized in a Power BI report

We have been having issues with long refresh times (it can take around 4 hours to get the data from Snowflake to the report). We have tried implementing incremental refresh but have run into issues since we need to consider deletes and due the how the incremental refresh partitions work, this then creates duplicates which take too long to remove using Power Query.

Is there a way to rework the architecture using the new Fabric capabilities so it performs better?

We are on premium capacity.

Thank you for any help!

[ibarrau]

Hi. The answer would be "It depends". It's not necessary an issue about the tech, but the decisions of the data transformations. How are you using dataflow? what kind of data transformations (joins, append, changing types, etc) are you doing? how many rows? how many columns? Those questions might help us understand if there is an issue about practices or if it is the tech.

I have worked with small millions of data three years ago, without dataflows and worked ok. I mean Snowflake -> Power Bi Dataset -> Power Bi Report.

All transformations were applied at snowflake with queries creating dim and fact tables. The dataset would only read table.

I hope that helps