It has been a little over a month since Microsoft announced Microsoft Fabric, with the following objectives:
Unify your data estate
Establish an open and lake-centric hub that helps data engineers connect and curate data from different sources—eliminating sprawl and creating custom views for everyone.
Manage powerful AI models
Accelerate analysis by developing AI models on a single foundation without data movement—reducing the time data scientists need to deliver value.
Empower everyone in your business
Innovate faster by helping every person in your organization act on insights from within Microsoft 365 apps, such as Microsoft Excel and Microsoft Teams.
Govern data across your organization
Responsibly connect people and data using an open and scalable solution that gives data stewards additional control with built-in security, governance, and compliance.
All your data. All your teams. All in one place.https://www.microsoft.com/en-us/microsoft-fabric
You probably recognize a similar ambition to Azure Synapse Analytics, don’t you? After trying to integrate Power BI into Synapse Analytics, Microsoft realized that converging a PaaS data platform by capitalizing on its historical engines (ADF, SQLDW, ADLS Gen 2) would not be enough in terms of usage or financial efficiency for its customers.
That’s why with Microsoft Fabric, the paradigm changes completely. The goal is to integrate all personas (renamed experiences for the occasion) into a SaaS solution like Power BI.
I find Microsoft’s choice extremely wise, focusing on experiences rather than the technological stack. Thus, with the same Fabric or Premium capacity, it becomes possible to execute a multitude of workloads. From data lake to reports, through ML models. It’s extremely ambitious, but the promise is fantastic.
Of course, this comes with a set of disadvantages that will undoubtedly make the solution unsuitable for certain business contexts. In Microsoft Fabric, there is not only a technological solution, but also a way to use and implement its data analysis platform.
Some will say that basic services allow many implementations by slightly twisting the model, but should we do it? Microsoft offers a solution that highlights the concepts of centralized lake, Lakehouse, Datawarehouse, and Datamart. From this organization will result a market-recognized way of organizing its platform, a vocabulary, and practices.
For the sake of maintenance, semantics, and your business, stick to these recognized practices! Do not play the apprentice wizards who will want to use only the Lakehouse because you were doing that in your previous solution…
To get back to the subject, with Microsoft Fabric, Synapse Analytics becomes obsolete. While some terms and screens are similar, and integrations are possible (SQL Dedicated Pool), start thinking about migration, because like Azure Analysis Services, Synapse Analytics is going nowhere.
Don’t expect an easy migration path, Microsoft Fabric is a new tool, and you will have to review a large part of the implementations. And if you start wanting to diverge from the implementation approach inherent in Microsoft Fabric, remember, Azure Databricks, Azure SQLDB, and Azure Data Factory will continue to exist and evolve. Perhaps your happiness lies in these implementations rather than in the SaaS approach of Microsoft Fabric…
Many of you ask me my feelings about Microsoft Fabric, I have already shared it on two occasions:
In essence, I strongly believe in this product, it follows standards that I strongly believe in:
- An approach where the lake must be accessible to be valued and therefore a central entry point! And an easy way to access data (Lakehouse)
- The need to have access to Data Warehouses to implement Kimball’s precepts, dimensional modeling, historization, and business rule implementation
- The need to give the business more precise control through Datamarts with a very close and specific view of operational needs
- Native integration of all these layers for easy and secure use
- In the background, the advent of a real Analytics Engineer position rather than current Data Engineers who are valued only through technical knowledge (Spark, Python, Software) than by the ability to make data accessible and focused on the business (SQL, Dimensional Modeling, DAX).
However, Microsoft leaves us today in an uncomfortable situation. It is clear that Microsoft Fabric is the future of the managed data platform, and the service is currently in preview. My optimism is mainly based on the marketing promises of the build… It is now time to materialize these promises (One Security, DirectLake on DWH and DMT, integration at the level of ADF…) and announce the official release date of the solution!