Microsoft Fabric: OneLake Workspace Settings and Data Workflow (Airflow)

Microsoft Fabric unifies data engineering, analytics, storage, and orchestration into a single SaaS platform. For the DP-700: Microsoft Fabric Data Engineer Associate certification, understanding how OneLake workspace settings and Data Workflow (Airflow) work is critical.

These settings control:

• How files behave inside Fabric
• How external data is accessed
• How orchestration compute is managed
• How costs and performance are optimized

This guide walks through the key concepts you need for the DP-700 exam and real-world Fabric implementations.

OneLake Workspace Settings

OneLake is the central storage layer of Microsoft Fabric. It acts as a unified data lake across all Fabric experiences, including Lakehouses, Warehouses, and Notebooks.

Workspace settings define how storage behaves when interacting with local machines and external cloud storage systems.

OneLake File Explorer

OneLake File Explorer allows Fabric storage to appear like a local drive on your computer.

This makes it easier for engineers to work with files without directly interacting with cloud APIs.

Placeholders

Files appear locally but remain stored in the cloud.

A blue cloud icon indicates the file is only a placeholder.

When you open the file:

• Fabric downloads the file from OneLake
• The icon changes to a green check mark

This approach saves disk space while still providing seamless access to cloud data.

Sync Behavior

Synchronization works differently from typical cloud storage systems.

Local Changes

When you modify a file locally, it automatically uploads to OneLake.

Cloud Changes

Changes made in Fabric or by other users do not automatically sync to your local machine.

You must manually run:

Right Click → Sync from OneLake

This prevents unexpected overwrites during collaborative development.

Important distinction: OneLake File Explorer behaves differently from tools like OneDrive. OneDrive is sync-heavy, meaning files continuously synchronize between local and cloud storage. OneLake, however, is designed to be on-demand, where files are downloaded only when accessed and cloud updates require manual synchronization. This difference is a common point of confusion and often appears as an exam distractor in DP-700 questions.

Case Sensitivity

OneLake storage is case sensitive, but Windows file systems are not.

Example:

file.txt File.txt

Both files can exist inside OneLake.

However, Windows File Explorer may only display the oldest file, which can cause confusion when working locally.

OneLake Shortcuts

Shortcuts allow Fabric to reference data stored in external systems without copying it.

Instead of moving data into OneLake, Fabric creates a logical pointer to the data location.

Supported external sources include:

• Azure Data Lake Storage
• Amazon S3
• Google Cloud Storage
• Other Fabric workspaces

This enables a data virtualization approach, reducing storage duplication.

Tables vs Files

Shortcuts behave differently depending on whether they reference tables or files.

Tables

Table shortcuts must follow strict rules.

They must:

• Exist at the top level of the Lakehouse
• Avoid spaces in folder names
• Follow the Delta Lake structure

When these conditions are met, Fabric can automatically discover the table.

Example structure:

Tables/sales_data

Fabric automatically registers this as a Delta table.

Files

File shortcuts are more flexible.

They:

• Can exist inside nested folders
• Do not automatically register as tables
• Must be manually loaded into Spark or SQL queries

Example structure:

Files/raw/2025/january

These are treated as unstructured storage.

OneLake Permissions

Fabric supports two access patterns when reading external data.

Internal Access

Internal access uses the identity of the user running the query.

Flow:

User → Fabric → External Storage

The external storage system verifies the user’s credentials.

External Access

External access uses the connection identity created by the shortcut owner.

Flow:

User → Fabric → Shortcut Owner Connection → Storage

This allows organizations to control access without exposing external storage credentials to all users.

For DP-700, a simple way to remember this distinction is:

Internal Access: “I am who I am” — the user’s Microsoft Entra ID identity is passed directly to the external system (identity passthrough).

External Access: “I am using the shortcut creator’s key” — Fabric uses the credentials defined in the shortcut connection, such as a Service Principal or key-based authentication.

OneLake Caching

Caching helps reduce latency and cloud egress costs when accessing external data.

Caching applies to external storage systems such as:

• Amazon S3
• Google Cloud Storage
• On-premise storage

Key behavior:

• Files smaller than 1 GB may be cached
• Cache duration ranges between 1 and 28 days

Benefits include:

• Faster queries
• Lower external storage costs
• Reduced network overhead

Delegated Identity

Some Fabric engines do not use the end-user identity when accessing external data.

Instead, they use the Item Owner’s permissions.

This behavior applies to:

• T-SQL queries
• Direct Lake mode

Example flow:

User runs query → Fabric uses Item Owner identity → External storage accessed

This ensures consistent access control when multiple users query shared datasets.

This behavior is especially important when using Direct Lake mode in Power BI. When a Power BI report reads data directly from OneLake using Direct Lake, the query typically runs using the identity of the Lakehouse item owner, not the user viewing the report. This design ensures consistent access control across shared semantic models and is an important security concept in Microsoft Fabric.

Data Workflow (Airflow) in Microsoft Fabric

Microsoft Fabric includes Data Workflow, which is built on Apache Airflow.

Airflow enables orchestration of:

• Spark jobs
• Notebooks
• Pipelines
• Data movement tasks
• External service integrations

Workflows run on compute pools configured in workspace settings.

Starter Pool

Starter pools are designed for development and testing workloads.

Key characteristics:

• Automatically shuts down after 20 minutes of inactivity
• Starts instantly when active
• Takes about 5 minutes to resume if paused

Starter pools are ideal for:

• Pipeline experimentation
• Notebook testing
• Development orchestration

They provide lower cost compute for non-production scenarios.

Custom Pool

Custom pools are designed for production workloads.

Unlike starter pools, custom pools remain active and support scaling.

Key capabilities:

• Always running compute
• Autoscaling support
• Parallel task execution

Each additional node increases worker capacity.

Example:

1 node → 3 workers 2 nodes → 6 workers 3 nodes → 9 workers

This allows multiple Airflow tasks to run in parallel, improving workflow performance.

In most standard Fabric workspaces, the maximum node count for a Custom Pool is typically 10 nodes. Since each node contributes three workers, this results in a maximum of about 30 concurrent workers for executing Airflow tasks.

Networking Limitations

An important detail for DP-700:

Currently, Data Workflow in Microsoft Fabric does not support:

• Virtual Networks (VNet)
• Private Link connectivity

All workflow interactions occur through public endpoints.

Additional Concepts Important for DP-700

Understanding how OneLake organizes data is also useful for the exam.

Each workspace can contain:

• Lakehouses
• Warehouses
• KQL Databases
• Notebooks
• Pipelines
• Dataflows

However, all storage ultimately resides inside OneLake.

Structured data inside Lakehouses is typically stored using Delta Lake format, which provides:

• ACID transactions
• Schema enforcement
• Time travel
• Optimized Spark performance

Key Takeaways for the DP-700 Exam

To succeed in DP-700, you should clearly understand:

OneLake

• File placeholders and sync behavior
• Case sensitivity differences
• Table vs file shortcuts
• Internal vs external permissions
• Caching behavior
• Delegated identity

Data Workflow

• Starter pools vs custom pools
• Autoscaling workers
• Airflow orchestration model
• Networking limitations

Final Thoughts

Microsoft Fabric simplifies modern data architectures by integrating storage, compute, orchestration, and governance into a single platform.

Understanding OneLake behavior and Data Workflow orchestration is essential not only for passing the DP-700 certification exam, but also for building scalable and cost-efficient data engineering solutions in Fabric.

Mastering these workspace settings will help you design production-ready pipelines, optimize performance, and manage data securely across cloud environments.

References

OneLake File Explorer https://learn.microsoft.com/en-us/fabric/onelake/onelake-file-explorer

OneLake Shortcuts and Caching https://learn.microsoft.com/en-gb/fabric/onelake/onelake-shortcuts#caching

Introducing Data Workflows in Microsoft Fabric https://blog.fabric.microsoft.com/en-us/blog/introducing-data-workflows-in-microsoft-fabric?ft=All

Apache Airflow Job Workspace Settings https://learn.microsoft.com/en-us/fabric/data-factory/apache-airflow-jobs-workspace-settings

Thanks for Reading

Thank you for taking the time to read this guide.

If you found this helpful while preparing for DP-700 or working with Microsoft Fabric, feel free to share it with others who might benefit.

If you notice any mistakes, missing details, or updates, please leave a comment so the article can continue improving for everyone.