Lakehouse-Architektur

Last modified by Dimitri Rupp on 2026/06/11 09:30

Contents

💧 Fabric Lakehouse

💧 Fabric Lakehouse

A Lakehouse in Microsoft Fabric is a unified data architecture that merges the scalability of a data lake with the structure and performance of a data warehouse, offering a single platform for storing, managing, and analyzing structured and unstructured data.

🗄️ Storage

Data is stored in Delta tables, which support ACID transactions, schema evolution, and time travel. A Lakehouse contains two zones — Tables and Files.

Tables — managed Delta tables with full metadata support
Files — unmanaged storage for raw data (CSV, Parquet, images, etc.), accessible via Spark but not queryable via SQL endpoint unless converted to Delta

🔄 Transformation

Dataflows Gen2, Notebooks (PySpark / Spark SQL), and Pipelines enable ingestion, cleansing, and transformation. These tools integrate natively with Lakehouses for modular, reproducible workflows.

🔐 Access

Lakehouses expose a SQL endpoint for T-SQL queries. Access is governed via Microsoft Entra ID, with support for workspace roles, item-level permissions, and Service Principals for secure external automation.

🏗️ Layers in our architecture

The SPL Fabric architecture follows the medallion pattern with three layers:

flowchart LR
    SRC[📂 Source Systems] --> B[🥉 Bronze
Landing zone]
    B --> S[🥈 Silver
Transformation]
    S --> G[🥇 Gold
Business-ready]
    G --> CONS[📊 Consumers
Power BI / Excel / Board]

    style B fill:#cd7f32,color:#fff
    style S fill:#c0c0c0,color:#000
    style G fill:#ffd700,color:#000

Bronze Lakehouse — landing zone with raw data from source systems
Silver Lakehouse — transformation zone, cleaned and enriched data
Gold Lakehouse — business-ready datasets for reporting and analytics

For detailed access instructions, see Get data from Lakehouse.

📋 Source-specific Bronze Lakehouses

The Bronze layer is split into individual Lakehouses per source system:

💻 BMD Bronze Lakehouse — data from BMD
🗄️ Rimo Lakehouse — data from Rimo DWH
🔧 Board Bronze Lakehouse — data from Board (via Gen2 Dataflow)
🏗️ H3A Bronze Lakehouse — data from H3A
📁 SharePoint Bronze Lakehouse — data from SharePoint files