Cloud Data Platforms Data Engineering Microsoft Fabric

Modern Data Warehouse Architecture with Microsoft Fabric

Modern Data Warehouse Architecture with Microsoft Fabric
Data Engineering

Modern Data Warehouse Architecture Microsoft Fabric: The Executive Guide

⏱️ 8 min read
Data Engineering · Microsoft Fabric
Modern data warehouse architecture diagram showing Microsoft Fabric Lakehouse Warehouse and Medallion layers with OneLake — Numlytics

Microsoft Fabric unifies the modern data warehouse into a single platform — replacing the fragmented stack of Azure Synapse, Data Factory, and Power BI Premium with one governed, consumption-based service

The concept of a modern data warehouse has changed significantly since Microsoft Fabric reached General Availability in late . What was previously a complex stack of separate Azure services Synapse Analytics for the warehouse, Data Factory for pipelines, Power BI Premium for reporting, and Azure Data Lake for storage — is now unified under a single platform with a single governance model, a single storage layer (OneLake), and a single capacity SKU. For data leaders who last evaluated their data architecture two or three years ago, the options available today are materially different from what they remember.

This guide explains the primary modern data warehouse architecture patterns available in Microsoft Fabric in , how to choose between them, and what the architectural decision means for governance, cost, and the speed at which your analytics team can deliver value.

What Changed: Why the Data Warehouse Looks Different

Three shifts have fundamentally changed the architecture decisions available to enterprise data teams. First, the separation between data lake and data warehouse has collapsed. Microsoft Fabric's OneLake stores all data in Delta Parquet format, meaning the same physical data can be queried by a Fabric Warehouse (SQL-first), a Fabric Lakehouse (Spark-first), or Power BI (DAX-first) without duplication or transformation. The architectural choice is now about the processing layer, not the storage layer.

Second, the cost model has shifted from reserved capacity to consumption-based billing. Azure Synapse required significant upfront capacity commitment. Fabric's F-SKU pricing can be paused, scaled, and right-sized dynamically — changing the economics of running a data warehouse for organisations with variable workload patterns.

Third, the governance layer is now built in. Microsoft Purview integration, sensitivity labels, and data lineage tracking are native to Fabric — not bolt-on additions requiring separate configuration and licensing. This materially reduces the effort and cost of meeting data governance and compliance requirements.

"In 2025, a modern data warehouse is not a piece of infrastructure you build and maintain — it is a governed, unified platform you configure and consume. Microsoft Fabric is the clearest expression of this shift in the Microsoft ecosystem."

Microsoft Fabric as the Modern Data Warehouse Platform

Microsoft Fabric is an end-to-end analytics platform that integrates data engineering, data integration, data warehousing, real-time analytics, and business intelligence into a single SaaS offering. All data is stored in OneLake — a single, tenant-wide data lake built on Azure Data Lake Storage Gen2 — and all Fabric workloads (Lakehouse, Warehouse, Dataflows, Notebooks, Power BI) operate against this shared storage without copying data between services.

For organisations currently running Azure Synapse, Azure Data Factory, and Power BI Premium as separate services, migrating to Microsoft Fabric consolidates those workloads into a single capacity SKU, eliminates cross-service data movement, and provides unified governance through a single administrative console.

The Three Primary Architecture Patterns in Fabric

PatternBest ForPrimary Fabric ComponentQuery Language
Lakehouse-firstData engineering heavy workloads, ML, raw data explorationFabric Lakehouse + NotebooksPySpark / SQL
Warehouse-firstSQL-centric teams, traditional BI, governed reportingFabric WarehouseT-SQL
Hybrid MedallionEnterprise-scale, multiple consumers, full governanceLakehouse (Bronze/Silver) + Warehouse (Gold)PySpark + T-SQL

The Medallion Architecture: Bronze, Silver, Gold Explained

The Medallion architecture is the recommended pattern for enterprise Microsoft Fabric deployments and is Microsoft's own stated best practice for Fabric data engineering. It organises data into three logical layers, each with a distinct purpose and governance level.

Bronze layer — raw ingestion

The Bronze layer stores data exactly as it arrives from source systems — unmodified, timestamped, and retained indefinitely. It serves as the audit-grade record of everything the platform has received. In Fabric, this is typically implemented as a Lakehouse with Delta tables, and data is loaded via Data Factory pipelines or Dataflow Gen2.

Silver layer — cleansed and conformed

The Silver layer applies business rules, data quality checks, and schema conformance to produce a clean, standardised dataset. Duplicate records are resolved, nulls are handled, and reference data is joined. This layer is the single source of truth for each domain entity — customers, products, transactions — before any business-specific aggregation is applied.

Gold layer — business-ready analytics

The Gold layer contains pre-aggregated, business-logic-enriched datasets designed for direct consumption by Power BI semantic models and self-service analytics. In Fabric, this is typically implemented as a Fabric Warehouse or a Lakehouse SQL endpoint, enabling T-SQL access for Power BI Direct Lake connections — the fastest possible query path for Power BI reports against Fabric data.

Fabric Lakehouse vs Fabric Warehouse: How to Choose

DimensionFabric LakehouseFabric Warehouse
Primary query languagePySpark (Notebooks) + SQL endpointT-SQL (full DML/DDL support)
Best forData engineering, ML, raw data, semi-structured dataGoverned reporting, SQL-centric BI teams
TransactionsAppend-optimised (Delta)Full ACID transactions
Power BI connectivity✓ Direct Lake (fastest)✓ Direct Query or Direct Lake via Mirroring
Multi-table updatesVia Spark or SQL endpoint (limited)✓ Full UPDATE/DELETE/MERGE
Team skill requirementPython / Spark experienceSQL experience (accessible to more teams)

Microsoft Fabric vs Azure Synapse: Is Migration Worth It?

For organisations currently running Azure Synapse Analytics, the migration question is increasingly pressing. Microsoft has not deprecated Synapse, but the direction of investment is clearly Fabric. New features — Direct Lake, Copilot for data engineering, real-time analytics, and unified governance — are being developed exclusively for Fabric. Synapse will continue to receive security updates and support, but organisations building new capabilities on Synapse are investing in a platform that will become progressively less differentiated.

The migration case is strongest for organisations that are also using Power BI Premium or Azure Data Factory, since consolidating all three workloads into Fabric typically produces a net cost reduction while improving performance and governance. Numlytics has delivered Microsoft Fabric migration programmes for organisations across the US, UK, Australia, and UAE — contact us for a migration readiness assessment specific to your current architecture.

Key Takeaways
  • Microsoft Fabric unifies data engineering, warehousing, and BI into a single platform with shared storage (OneLake) — eliminating the multi-service complexity of the previous Azure stack.
  • The Medallion architecture (Bronze → Silver → Gold) is Microsoft's recommended pattern for enterprise Fabric deployments and delivers the clearest governance and performance outcomes.
  • Choose Fabric Warehouse if your team is SQL-centric and primarily serves governed BI reporting. Choose Fabric Lakehouse if you have Spark expertise and need to process raw or semi-structured data at scale.
  • Power BI Direct Lake — available only in Fabric — delivers the fastest possible query performance for Power BI reports without the limitations of Import mode or the latency of DirectQuery.
  • Organisations running Azure Synapse + Data Factory + Power BI Premium should evaluate Fabric consolidation — the cost and governance case is compelling in most enterprise scenarios.

Next Steps: Designing Your Modern Data Architecture

The right modern data warehouse architecture for your organisation depends on your current data estate, team skills, source system landscape, and reporting requirements. There is no universal answer — but the architectural decision made today will determine the cost, flexibility, and governance quality of your analytics platform for the next three to five years.

Numlytics offers a structured Microsoft Fabric architecture assessment that evaluates your current state, maps your requirements to the appropriate Fabric patterns, and produces a phased implementation roadmap with cost modelling. Speak with a certified Fabric consultant to start the conversation.