Skip to content

Microsoft Fabric Analytics Overview

Overview

Microsoft Fabric is a unified analytics platform that brings together data integration, data engineering, data warehousing, data science, real-time analytics, and business intelligence.

Fabric Components

OneLake

  • Unified data lake for entire organization
  • Hierarchical namespace
  • Parquet and Delta Lake format
  • Single copy of data, multiple workloads

Data Factory (Fabric)

  • Data integration and orchestration
  • Dataflows Gen2: Power Query at scale
  • Data pipelines: Copy and transform data
  • Similar to Azure Data Factory

Synapse Data Engineering

  • Apache Spark for data engineering
  • Notebooks: Interactive development
  • Spark job definitions
  • Lakehouse: Data storage and compute

Synapse Data Warehouse

  • Enterprise data warehouse
  • T-SQL queries
  • Auto-tune and optimize
  • Separates storage and compute

Synapse Real-Time Analytics

  • Event streaming and analysis
  • Kusto Query Language (KQL)
  • Time-series data
  • IoT and log analytics

Power BI

  • Business intelligence and reporting
  • Direct Lake mode: Query OneLake directly
  • Interactive dashboards
  • Natural language queries (Q&A)

Data Science

  • Machine learning workflows
  • Notebooks: Python, R, Scala
  • MLflow integration
  • AutoML capabilities

Lakehouse Architecture

What is a Lakehouse?

  • Combines: Data lake + Data warehouse
  • Delta Lake format (ACID transactions)
  • Schema enforcement
  • Time travel (versioning)

Files vs Tables

Files: - Unstructured data - Any format: CSV, JSON, Parquet - Flexible schema-on-read

Tables: - Structured data - Delta Lake format - ACID transactions - Query with T-SQL or Spark SQL

Semantic Models

Creating Semantic Models

  • Data modeling layer for Power BI
  • Define relationships between tables
  • Create measures and calculations (DAX)
  • Reusable across reports

Direct Lake Mode

  • Query OneLake directly
  • No data import needed
  • Best performance
  • Automatic refresh

Data Integration

Dataflows Gen2

  • Power Query engine
  • Transform data at scale
  • Output to Lakehouse, Warehouse, or KQL DB
  • Incremental refresh

Data Pipelines

  • Copy activity for data movement
  • Transform with notebooks or Spark
  • Schedule and orchestrate
  • Integration with Azure services

Real-Time Analytics

Eventhouse and KQL Databases

  • Event streaming
  • Time-series data
  • Real-time dashboards
  • IoT scenarios

Streaming Data

  • Event Streams: Kafka-compatible
  • Real-time processing
  • Integration with Fabric components

Security and Governance

Workspace Roles

  • Admin: Full control
  • Member: Create and edit
  • Contributor: Edit existing
  • Viewer: Read-only access

Data Access

  • OneLake data access roles
  • Row-level security (RLS) in semantic models
  • Object-level security (OLS)

Monitoring

  • Monitoring hub: Track activities
  • Metrics: Performance and usage
  • Alerts: Proactive notifications

Best Practices

Lakehouse Design

  1. Medallion architecture: Bronze → Silver → Gold
  2. Use Delta Lake for tables
  3. Partition large tables appropriately
  4. Optimize file sizes (128MB-1GB)

Semantic Models

  1. Star schema design (fact + dimension tables)
  2. Use Direct Lake when possible
  3. Create calculated columns and measures in DAX
  4. Implement RLS for security

Performance

  1. Optimize Spark cluster sizes
  2. Use V-Order for faster reads
  3. Partition and Z-order data
  4. Cache frequently accessed data

Study Tips

Key Concepts

  • OneLake as unified data lake
  • Lakehouse vs Warehouse vs KQL DB
  • Semantic models and Direct Lake mode
  • Dataflows vs Data Pipelines
  • Medallion architecture (Bronze/Silver/Gold)

Common Scenarios

  1. Data engineering → Lakehouse with Spark
  2. Data warehousing → Synapse Data Warehouse
  3. Real-time analytics → KQL Database
  4. BI and reporting → Power BI with semantic models
  5. ML workflows → Data Science notebooks

Remember

  • Fabric = All-in-one analytics platform
  • OneLake = Unified data lake
  • Lakehouse = Lake + Warehouse benefits
  • Direct Lake = Fastest Power BI mode
  • Medallion = Bronze (raw) → Silver (cleaned) → Gold (curated)