Microsoft Fabric: The Unified Analytics Platform

Microsoft Fabric (generally available since November 2023) is Microsoft’s answer to the fragmentation of the modern data stack. Rather than stitching together Azure Synapse, Azure Data Factory, Azure Data Lake, Power BI, and Azure Machine Learning separately, Fabric provides a single, unified SaaS analytics platform that covers the entire analytics lifecycle.

Overview

Fabric converges six workloads into one integrated experience, built on a foundation of OneLake (a single, tenant-wide data lake based on ADLS Gen2) and open formats (Delta Lake, Parquet).

Core philosophy: One lake. One platform. One billing. Zero data movement between tools.

The Six Fabric Workloads

┌─────────────────────────────────────────────────────────────────┐
│                    MICROSOFT FABRIC                              │
│                                                                  │
│  ┌─────────────┐  ┌───────────┐  ┌──────────────────────────┐  │
│  │  Data       │  │  Data     │  │  Data                    │  │
│  │  Engineering│  │  Factory  │  │  Warehouse               │  │
│  └─────────────┘  └───────────┘  └──────────────────────────┘  │
│  ┌─────────────┐  ┌───────────┐  ┌──────────────────────────┐  │
│  │  Real-Time  │  │  Data     │  │  Power BI                │  │
│  │  Analytics  │  │  Science  │  │  (BI & Reporting)        │  │
│  └─────────────┘  └───────────┘  └──────────────────────────┘  │
│                                                                  │
│  ──────────────────── ONELAKE ─────────────────────────────────  │
│            (Single Tenant Data Lake - ADLS Gen2)                │
└─────────────────────────────────────────────────────────────────┘

1. Data Engineering (Lakehouse)

A Spark-based Lakehouse environment with:

• Notebooks (Python, Scala, Spark SQL, R)
• Spark Job Definitions for scheduled batch workloads
• Data stored as Delta Lake tables in OneLake
• Auto-provisioned Spark clusters — no cluster management needed

2. Data Factory

Low-code/no-code data integration with:

• Pipelines: orchestrate complex data workflows (drag-and-drop)
• Dataflows Gen2: Power Query-based transformations at scale
• 200+ native connectors to SaaS apps, databases, and files
• Integration with Azure Data Factory and Azure Integration Runtime

3. Data Warehouse

A fully managed, enterprise-grade SQL warehouse with:

• T-SQL support with full DML (INSERT, UPDATE, DELETE, MERGE)
• Distributed query execution on Delta/Parquet tables
• Automatic statistics management and query optimisation
• Row/column-level security; dynamic data masking
• Storage in OneLake as Parquet files — open, queryable from outside Fabric

4. Real-Time Analytics (KQL Database)

A high-performance time-series and event analytics engine powered by the Kusto Query Language (KQL):

• Ingests millions of events per second from EventStream
• Sub-second query latency on terabytes of time-series data
• Built-in dashboards, anomaly detection, and forecasting
• EventStream: managed streaming ingestion from Kafka, Event Hubs, IoT Hub

5. Data Science

Integrated ML development environment with:

• Notebooks with built-in MLflow experiment tracking
• Models versioning and management
• Integration with Azure ML for production deployments
• Direct access to Lakehouse Delta tables for training
• AutoML and built-in AI functions (e.g. PREDICT() in SQL)

6. Power BI

The industry-leading BI platform, now natively embedded in Fabric:

• Direct Lake mode: Power BI reads Delta/Parquet files directly from OneLake — no data import, no DirectQuery overhead — the fastest BI connectivity available
• Report authoring, data modelling, and administration in one place
• Real-time dashboard updates via streaming datasets

OneLake: The Foundation

OneLake is the single, unified data lake that underpins all Fabric workloads:

Shortcuts

A powerful OneLake feature that creates virtual references to data in other storage systems:

• Point to S3 buckets, ADLS Gen2 accounts, or Google GCS paths
• Data is not copied — queries are executed against the source
• Enables multi-cloud and hybrid scenarios without data movement

Fabric Capacity & Licensing

Fabric uses a capacity-based model (unlike per-user or per-query pricing):

All workloads within a capacity share the same compute units. Power BI Premium per capacity users get Fabric included at the same SKU.

Key Differentiators vs. Azure Synapse

Strengths

Limitations

When to Choose Microsoft Fabric

✅ Good fit when:

• Your organisation is Microsoft/Azure-first (M365, Teams, Azure AD)
• You want to consolidate multiple Azure data services into one platform
• Power BI is your BI tool and you want the fastest possible connectivity (Direct Lake)
• You need unified governance across engineering, analytics, and science workloads
• You want a managed SaaS experience without infrastructure overhead

❌ Poor fit when:

• You are multi-cloud (AWS/GCP primary) and can’t commit to Azure-centric OneLake
• You need extreme Spark customisation (Fabric’s managed Spark has less flexibility than Databricks)
• You are already deeply invested in Databricks or Snowflake with mature pipelines
• Your use case is primarily streaming at very high throughput (dedicated Kafka + Flink may outperform)