Skip to content

Organize a Fabric Lakehouse using Medallion Architecture Design

Organize a Fabric Lakehouse using Medallion Architecture Design

Source

Summary

This module establishes the enterprise design principles of the Medallion Architecture (Bronze $\rightarrow$ Silver $\rightarrow$ Gold) within Microsoft Fabric Lakehouses to incrementally clean, curate, validate, and enrich data assets for high-performance consumption.

Medallion Architecture Breakdown

[External Sources]
▼ (Append-only / Raw ingestion)
┌────────────────────────────────────────────────────────┐
│ 🥉 BRONZE (Raw Zone) │
│ - Raw, unvalidated, historical audit trail │
│ - Formats: Parquet / Delta Lake │
│ - Ingestion: Data Factory Pipelines, Eventstreams │
└──────────────────────────┬─────────────────────────────┘
▼ (Cleaning, Deduplication, Schema Enforcement)
┌────────────────────────────────────────────────────────┐
│ 🥈 SILVER (Cleansed / Conformed Zone) │
│ - Standardized schema, normalized relationships │
│ - Deduplicated, cleansed strings, UTC timestamps │
│ - Engine: PySpark Notebooks / Dataflows Gen2 │
└──────────────────────────┬─────────────────────────────┘
▼ (Star Schema, Aggregations, Business Logic)
┌────────────────────────────────────────────────────────┐
│ 🥇 GOLD (Curated / Business Zone) │
│ - Star schema: Fact & Dimension tables │
│ - Optimized for Direct Lake reporting & Analytics │
│ - Engine: Fabric Data Warehouse / Lakehouse SQL │
└────────────────────────────────────────────────────────┘

Layer Characteristics in Microsoft Fabric

LayerPrimary RoleData StructureTypical Ingestion EngineTarget Consumers
BronzeRaw data preservationAppend-only raw Delta tables or filesData Factory Copy Activity, EventstreamsData Engineers, Reprocessing Pipelines
SilverCleaned enterprise viewConformed, deduplicated Delta tablesPySpark Notebooks, Dataflows Gen2Data Engineers, Data Scientists
GoldBusiness aggregations & BIStar Schema (Dimensions & Facts)PySpark Notebooks, Fabric Data WarehousePower BI, Business Analysts, Executive Dashboards

Multi-Workspace vs Single-Lakehouse Topology

  1. Single Lakehouse with Layer Schemas: Low complexity, suitable for small to mid-sized analytical projects.
  2. Multi-Lakehouse (Bronze LH, Silver LH, Gold LH): Enhanced governance, distinct workspace security boundaries, independent compute/storage scaling.

Exam Traps & Gotchas

[!IMPORTANT]

  • Bronze is Append-Only: Never perform destructive schema mutations or updates in the Bronze layer; it serves as the ultimate historical recovery source.
  • Direct Lake Compatibility: Ensure Gold layer tables are fully structured Delta tables with valid relationships and without complex calculated DAX columns on the model side, enabling full Direct Lake acceleration.