5.2.1. OneLake: The Foundation
💡 First Principle: OneLake is a single, logical data lake for your entire organization. Think of it as "OneDrive for data"—one namespace, automatic data organization, built-in governance. All Fabric workloads read from and write to OneLake, eliminating data silos.
OneLake Key Features
- Single namespace: All data appears under one logical hierarchy regardless of physical location
- Shortcuts: Virtual pointers to external data (AWS S3, ADLS Gen2, Google Cloud Storage) without copying
- Open format: Data stored as Delta Parquet by default—open, interoperable, no lock-in
- Built-in discovery: every workload sees the same namespace, so data is findable across the tenant without a separate catalogue step
Shortcuts: Access Without Copying
A shortcut is a pointer that makes data stored elsewhere appear inside OneLake as if it lived there. Point one at ADLS Gen2, Amazon S3, Google Cloud Storage or another Fabric workspace, and every Fabric workload can read it in place.
This is the feature that makes the "one copy" claim real. In a traditional architecture the same table is copied into the lake, then into the warehouse, then imported into the BI cache — three copies to govern, secure and keep in step. A shortcut has no copy to drift.
Why Delta Parquet Matters
OneLake stores tables as Delta Parquet by default: Parquet's columnar layout plus the Delta transaction log. That single choice is what lets a SQL warehouse query and a Spark notebook write the same table safely, because Delta adds ACID transactions and versioning on top of files in a lake.
Because the format is open, data written by Fabric is readable by anything that understands Delta — there is no proprietary lock-in.
⚠️ Exam Trap: OneLake is not "another data lake service" competing with Data Lake Storage Gen2 — it is built on Gen2 and there is exactly one OneLake per tenant, provisioned automatically. You do not create OneLake accounts the way you create storage accounts. If a question asks you to provision storage per project, it is describing Gen2; if it describes a single organisation-wide namespace every workload shares, that is OneLake.
