Episode Details
Back to Episodes
Microsoft Fabric Data Platform: Why It’s Becoming the New Operating System for Enterprise Data
Season 1
Published 4 months, 2 weeks ago
Description
(00:00:00) The Fabric Platform: A Unified Approach to Data Management
(00:00:45) The Fragmentation Problem
(00:02:24) Fabric: A Solution to Fragmentation
(00:04:37) The Medallion Architecture
(00:09:07) Direct Lake and Semantic Models
(00:17:30) Workspaces and Security
(00:23:44) Edge Cases and Real-Time Operations
(00:28:18) Hybrid Walkthrough: One Lake and Purview Security
(00:35:59) Seven-Day Implementation Plan
(00:42:36) The Fabric Mindset Shift
In this episode of M365.fm, Mirko Peters explains why you don’t have a real data platform today—just a staged illusion held together by Power BI and pipelines—and how Microsoft Fabric, OneLake, and Medallion turn that chaos into a single, auditable enterprise data OS.
WHAT YOU WILL LEARN
You don’t have a platform if you can’t name your access path, your contracts, and your single place of truth. You have sprawl with dashboards on top.
Fabric exists to attack fragmentation: one identity (Entra), one storage layer (OneLake), one governance plane (Purview + workspaces), one monitoring view for warehouses, pipelines, notebooks, and reports.
Medallion only works when Bronze is evidence, Silver is truth, and Gold is meaning—and when each layer has clear owners, tests, and blast radius limits.
This episode argues that Fabric is not “one more tool,” but the moment you compress your surface area so there are simply fewer places to lie.
WHY MICROSOFT FABRIC AS DATA OS WORKS
(00:00:45) The Fragmentation Problem
(00:02:24) Fabric: A Solution to Fragmentation
(00:04:37) The Medallion Architecture
(00:09:07) Direct Lake and Semantic Models
(00:17:30) Workspaces and Security
(00:23:44) Edge Cases and Real-Time Operations
(00:28:18) Hybrid Walkthrough: One Lake and Purview Security
(00:35:59) Seven-Day Implementation Plan
(00:42:36) The Fabric Mindset Shift
In this episode of M365.fm, Mirko Peters explains why you don’t have a real data platform today—just a staged illusion held together by Power BI and pipelines—and how Microsoft Fabric, OneLake, and Medallion turn that chaos into a single, auditable enterprise data OS.
WHAT YOU WILL LEARN
- Why most “modern data stacks” are really copy storms, shadow truths, and governance theater
- How to stop using Power BI as duct tape and design a single access path from raw → insight
- How to make Bronze / Silver / Gold real contracts instead of slideware
- How Fabric, OneLake, Purview, and workspaces work together to kill drift, silent copies, and unprovable numbers
- How Direct Lake changes Power BI by reading Delta in OneLake without imports or DirectQuery pain
- How to design multi‑workspace architecture so Platform owns Silver and domains own Gold
- How to stand in front of an executive and prove exactly where a number came from
You don’t have a platform if you can’t name your access path, your contracts, and your single place of truth. You have sprawl with dashboards on top.
Fabric exists to attack fragmentation: one identity (Entra), one storage layer (OneLake), one governance plane (Purview + workspaces), one monitoring view for warehouses, pipelines, notebooks, and reports.
Medallion only works when Bronze is evidence, Silver is truth, and Gold is meaning—and when each layer has clear owners, tests, and blast radius limits.
This episode argues that Fabric is not “one more tool,” but the moment you compress your surface area so there are simply fewer places to lie.
WHY MICROSOFT FABRIC AS DATA OS WORKS
- OneLake becomes the single organizational lake with open Delta/Parquet tables and shortcuts instead of copies
- All experiences (Data Factory, Engineering, Warehouse, Real‑Time, Data Science, Power BI, Data Activator) sit on the same storage, identity, governance, and monitoring plane
- Tables, not pipelines, become the contract, so schema drift and quality issues are visible and testable
- Direct Lake lets semantic models read Delta directly, avoiding import bloat and DirectQuery latency
- Multi‑workspace design (Platform vs domain vs shared analytics) brings clear ownership and promotion paths
- Cognitive load drops: fewer runtimes, fewer secrets, fewer “which thing runs where?” arguments
- If Power BI is acting as glue code, you don’t have BI, you have integration debt
- Bronze must stay messy and immutable, Silver must be validated and tested, Gold must be clean and business‑facing
- Shortcuts beat copy storms for connecting external stores into OneLake
- Platform teams should own shared Lakehouse and Silver; domains should own Gold and semant