OneLake: Understanding Fabric's Unified Storage Layer
OneLake is the single storage layer underpinning all of Microsoft Fabric. Here's how it differs from a traditional data lake and why it matters.
What is OneLake?
OneLake is Microsoft Fabric's unified storage layer — a single, governed data lake that serves as the foundation for all analytics workloads in the Fabric platform. Think of it as OneDrive for data: a centralised, hierarchical storage layer that every Fabric engine can access directly.
OneLake represents a significant architectural shift. Instead of each analytics engine maintaining its own storage silo, all Fabric workloads — SQL, Spark, real-time analytics, Power BI — read from and write to the same underlying data. This eliminates the data duplication, transformation pipelines and integration complexity that plague traditional multi-tool data stacks.
How OneLake works
OneLake is built on top of Azure Data Lake Storage Gen2 (ADLS Gen2) and uses Delta Lake as its default table format. Data is organised into workspaces, each containing shortcuts, data products and other artefacts.
Key concepts include:
- Workspaces — logical containers for data, artefacts and access control, analogous to folders in a file system
- Shortcuts — pointers to data stored elsewhere in OneLake or in external ADLS Gen2 accounts, enabling virtual data sharing without copying
- Delta tables — the default storage format, enabling ACID transactions, time travel and schema evolution across all Fabric engines
- Data products — governed, discoverable datasets that can be shared across workspaces and organisations
Why OneLake matters
Eliminating data duplication
In traditional architectures, data is often copied multiple times — from the data lake to the warehouse, from the warehouse to a semantic model, from the semantic model to a dashboard. Each copy adds storage cost, introduces freshness lag and creates governance complexity.
OneLake's shortcuts enable engines to access data in place. A Spark notebook, a SQL query and a Power BI report can all read the same Delta table without copying it. This reduces storage costs, ensures consistency and simplifies lineage.
Unified governance
Because all data lives in OneLake, governance is applied consistently. Access controls, sensitivity labels and lineage tracking work across all Fabric engines — there's no need to manage separate permission models for each tool.
DirectLake mode
OneLake enables Power BI's DirectLake mode, which reads Delta tables directly from OneLake without importing data or requiring a separate refresh schedule. This combines the performance of import mode with the freshness of direct query — a significant advancement for BI workloads.
OneLake and Databricks
OneLake is compatible with the open Delta Lake format, meaning Databricks can read and write to OneLake storage directly. This interoperability is strategically important — it enables organisations to use Databricks and Fabric together without duplicating data or maintaining separate storage layers.
Getting started
OneLake is automatically provisioned when you create a Fabric capacity — there's no separate setup. The key design decisions are around workspace structure, access management and how you'll organise shortcuts and data products.
Start with a clear workspace taxonomy, implement access controls early and take advantage of shortcuts to avoid data duplication. OneLake's value comes from treating it as a single, shared data estate — not as a collection of isolated workspaces.
