Calling Fabric Spark Notebooks from Power BI Paginated Reports

[Koraganti77]

Hi PBI Community,

I'm exploring a method to:

1. Call a Fabric Spark notebook from a Power BI paginated report
2. Process data in real-time using Spark
3. Return results to the paginated report for display

Key Questions:

• Has anyone implemented a similar Fabric Spark + PBI Paginated Reports integration?
• What APIs or methods could facilitate this connection?
• How to handle real-time data transfer between Spark and the report?

Use Case: Real-time financial reporting with complex calculations.

Challenges we're considering:

• Performance impact on report generation
• Data format compatibility
• Orchestrating the notebook execution

Any insights or experiences with similar integrations would be greatly appreciated!

Best

Srikanth K

[Anonymous]

Your ideas is so great lbendlin 

Hi, @Koraganti77 

If you need to connect to Fabric spark, you may need the following middleware to build a connection to your Power BI paginated report.

Supported data sources for Power BI paginated reports - Power BI | Microsoft Learn

You can select your data source as an input for the paginated report from the steps below:

If you want to process real-time data, once you've finished processing it in Spark, push it to your middle key, and the paginated report will read the latest data from it.

Performance depends on how you design your paginated report. You can refer to the following documents:

Troubleshoot report performance in Power BI - Power BI | Microsoft Learn

Design tips for reports in Power BI Report Builder - Power BI | Microsoft Learn

https://learn.microsoft.com/en-us/power-bi/guidance/report-paginated-data-retrieval

https://learn.microsoft.com/en-us/power-bi/paginated-reports/paginated-capacity-planning

Data compatibility, in terms of this, the data types that are currently supported by the Power BI semantic model. You can refer to the following links:

Report Definition Language (SSRS) - SQL Server Reporting Services (SSRS) | Microsoft Learn

Best Regards

Jianpeng Li

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[Koraganti77]

@Anonymous and @lbendlin 

Thank you for your insightful feedback. I appreciate your points about paginated reports and real-time data.

To provide some context: Currently, this process (real-time reporting) is happening via SSRS. The report triggers a Stored Procedure which performs all the complex ETL and sends the results back to the report.

We're in the process of migrating this platform to Fabric & Power BI, and replicating this real-time capability is part of that migration requirement. We're exploring how to best translate this SSRS/Stored Procedure approach into the Fabric ecosystem.

Given this background, do you have any thoughts on how we might approach this in Fabric and Power BI? Are there any features or patterns in these platforms that might help us achieve similar functionality to our current SSRS setup?

Thank you again for your help in thinking through this challenge.

Best

SK

[lbendlin]

I think we need to clarify what "real time" means to you. SSRS is not something I would associate with that as the RDLs need to be re-rendered.

Consider using Direct Query - but only if your data source is fast enough.

[Anonymous · Accepted solution]

Hi, @Koraganti77 

Thank you for your reply. Based on your description, I think you should be interested in Data Activator. It can automatically take actions when patterns or conditions are detected in the changed data. These actions will include ETL, such as triggering data flow to implement ETL operations, and then sending the latest data into the semantic model or warehouse.

Introduction to Data Activator - Microsoft Fabric | Microsoft Learn

What is Real-Time Intelligence - Microsoft Fabric | Microsoft Learn

Best Regards

Jianpeng Li

If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

[lbendlin]

In general paginated reports are pre-rendered, precisely because they take so long to render. They are the opposite of real time .

While technically possible, your approach will likely result in very unsatisfactory user experience.