Architecture Advice Needed : Manage historical Data
Hi everyone! 😊
We’re in the process of migrating our existing Azure SQL-based data warehouse to Microsoft Fabric, and I’d really appreciate some advice or suggestions from this amazing community. I’ve learned a lot from the posts and guidance shared here already, and I’m truly thankful for all the insights and support people offer.
Current Architecture:
  • We use Azure SQL DB as our data warehouse.
  • Data is staged from IBM DB2, then typed and incrementally loaded into SQL.
  • We have dim and fact views built on top, which are consumed by SSAS models and Power BI.
  • For historical data, we archive older views into tables (refreshed yearly), while the most recent 2 years remain live for easier access.
Planned Migration to Microsoft Fabric:
We're moving to a Lakehouse model in Fabric, with the following structure:
  • Bronze: Raw files stored as Parquet.
  • Silver: Cleaned and type-casted tables.
  • Gold: Business logic applied tables— targeting Direct Lake(not sure) for Power BI reporting.
Challenges and Questions:
Some of our business logic involves updates and merges, and we’re concerned about performance with large tables (some with 48 million rows). The business rules we need to apply include:
  • Converting all sales values to GBP
  • Finding and applying the latest product cost
  • Updating rebate percentages, which often change retroactively over the past 7–8 months
Also, we’re trying to figure out how to best handle historical data in Fabric — our current method of archiving into tables and keeping only recent data live (via views) has worked well.
What I’d love your input on:
  • What’s the best way to handle these kinds of business rules in Fabric without performance bottlenecks?
  • Should we move this logic to Dataflows Gen2, Notebooks, Lakehouse SQL, or something else?
  • Any best practices for managing historical data in Microsoft Fabric?
Any advice, war stories, or examples would be hugely appreciated! Thanks again to everyone who shares their knowledge here — it really helps people like me navigate these big changes. 🙏
1
1 comment
Sushmitha Heroor
2
Architecture Advice Needed : Manage historical Data
Learn Microsoft Fabric
skool.com/microsoft-fabric
Helping passionate analysts, data engineers, data scientists (& more) to advance their careers on the Microsoft Fabric platform.
Leaderboard (30-day)
Powered by