After a long and quite relaxing summer holiday Dark Matters is going strong. I’m quite busy with my own business development, resulting in a nice funnel already. My first confirmed assignment is regarding creating and providing a training for several Microsoft BI consultants covering all of Fabric, the new Microsoft SaaS BI Data Analytics offering.
And Fabric truly is a different beast compared to the previous offerings! The most notable stuff in my view relates to:
- Fabric is truly lake centric. It’s kinda funny that Fabric labels the two engines as Lakehouse (Spark) and Warehouse (SQL), but in line with the more common industry definition, both are lakehouses, as they use OneLake and Delta as the persistence layer. So study up on data lake design and Delta!
- Fabric admin and licensing is a nice mash-up of the new Fabric and the old Power BI licensing schemes. Also, we now have ‘domains’, an organizational container on top of workspaces. This allows for a Data Mesh approach, where each domain is more or less self-sufficient, and delivers production grade information products to end-users or other domains. The need for a central hub in the mesh, e.g. for analytical master data, seems likely for larger deployments.
- The Spark compute engine is the first-class citizen now in Fabric, trumping SQL. SQL will be there for a long time, but still, time to study up on Spark execution (and debugging and performance tuning) and PySpark (and all the cool 3rd party libraries, and how to manage these dependencies).
There is much more to cover regarding Fabric, so stay tuned, and share your thoughts!