Understanding Microsoft Fabric Architecture
Day2 of #FabricSeries
Microsoft describes Fabric as a unified, cloud-based data platform that covers the whole data lifecycle, from data ingestion to visualization.
Let’s break it down:
- End-to-end: You can ingest, store, transform, train, and visualize data in one environment.
- SaaS (software as a service): No servers to install or manage. Microsoft runs the infrastructure; you focus on data.
- Unified storage (OneLake): All workloads use the same logical data lake, so you avoid many copies of the same data.
- Shared experience: Data engineers, analysts, and scientists work in the same platform with shared items (lakehouse, warehouse, reports, etc.)
- Built-in governance: Permissions, sensitivity labels, and administration are applied consistently across Fabric items.
Microsoft Fabric Architecture: OneLake:
- All Fabric items (lakehouse, warehouse, Real-Time Analytics, Power BI Direct Lake datasets, etc.) store or read data through OneLake.
- OneLake can hold files (CSV, JSON, Parquet), structured tables, and unstructured data like logs or documents.
- You can access OneLake with APIs and tools that work with Azure Data Lake Storage (ADLS) and Azure Blob, just by using a OneLake URI.
Security and governance
- OneLake has centralized security and catalog features. The OneLake catalog helps you discover and govern data items, while OneLake security enforces role-based access on the data.
Compute & Workloads layer:
Data Engineering
- Use lakehouses to store both files and tables
- Use Spark notebooks and Spark job definitions to ingest, clean, and transform large volumes of data.
Data Warehouse
- SQL-based warehouse experience, optimized for analytical queries.
- Uses workload management to handle multiple users and queries efficiently.
Data Factory
- Orchestrates data movement and transformation at scale
- Supports batch pipelines and complex ETL/ELT
Real-Time Analytics
Handles streaming and time-series data for near real-time dashboards and alerts.
Data scientists work on the same platform, using notebooks and models on data in OneLake, and then share results with analysts.
Power BI
Analyst builds reports and dashboards
Because all these workloads share OneLake, the same data can be used by multiple teams without duplicate copies.
SaaS Services & Governance Layer:
- Uses Microsoft Entra ID (Azure AD) for authentication.
- Centralizes permissions and governance rules across items and data.
- Organizes Fabric content (lakehouses, warehouses, pipelines, reports, etc.) into workspaces and domains for better governance.
- The OneLake catalog lets users find and explore data items from a central place.
- Handles capacities, scaling, and monitoring so teams don’t manage servers or clusters directly. This layer makes Fabric feel like one product, even though it contains many engines and capabilities.
Thank you for reading! That’s a wrap for today
