Covers configuring Microsoft Fabric workspace settings, lifecycle management and governance, ingesting and transforming batch and streaming data, and monitoring and optimizing performance.
20 free sample questions167 in the full practice test Other versions: DP-201(74),DP-203(61)
A financial services firm is implementing a Microsoft Fabric solution to analyze trade data. They have a requirement to enforce data residency, ensuring that all data processing for their European operations occurs within the EU region. The Fabric capacity is provisioned in the 'West Europe' region. A data engineer creates a shortcut in a Lakehouse pointing to an Azure Data Lake Storage Gen2 account located in the 'East US' region. How will Fabric handle query processing against this shortcut?
Answer and explanation
Correct answer: C
Microsoft Fabric shortcuts do not move data; they are symbolic links. When a query is executed from a Fabric capacity in one region (e.g., 'West Europe') against a shortcut pointing to data in another region (e.g., 'East US'), the compute engine in the capacity's home region will pull the data across regions for processing. This results in cross-regional data egress and can violate data residency requirements if not managed properly. Fabric does not automatically block this action or move the data.
Question 2
You are designing a data ingestion pipeline for a large retail company. The pipeline must ingest nightly sales data from over 1,000 stores. Each store uploads a CSV file to an Azure Blob Storage container. You need to design a robust orchestration pattern in a Fabric pipeline that processes each file individually, logs its status, and can handle failures for a specific store's file without stopping the entire nightly batch. Which orchestration pattern should you implement?
Answer and explanation
Correct answer: B
This pattern is ideal for robust, item-level processing. The parent pipeline gets the list of files. The ForEach loop iterates through this list, passing each file name as a parameter to a child pipeline. This isolates the processing logic for each file, allowing for individual success/failure logging and retries within the child pipeline without affecting the processing of other files.
Question 3
You are optimizing a Fabric Warehouse that contains a 5-billion-row fact table named FactInternetSales. Queries against this table frequently filter by the OrderDateKey column. The data in the table is currently unordered. To improve query performance for time-series analysis, you decide to implement V-Order optimization. Which statement accurately describes the effect of applying V-Order on the OrderDateKey column?
Answer and explanation
Correct answer: B
V-Order is a write-time optimization for Delta Lake tables (which power the Fabric Warehouse) that physically sorts the data within the Parquet files. By ordering the data by a frequently filtered column like OrderDateKey, the query engine can perform file pruning (also known as data skipping) much more effectively, reading only the files that contain the relevant date ranges and significantly speeding up queries.
Question 4
A data engineering team is using Git integration with Azure DevOps to manage their Fabric workspace. They have two workspaces: Sales_Dev for development and Sales_Prod for production. A junior engineer commits a change to a notebook in the main branch directly from the Sales_Dev workspace. Now, they need to promote this change to the Sales_Prod workspace. The team uses a deployment pipeline for promotions. What is the immediate consequence of this action on the deployment pipeline process?
Answer and explanation
Correct answer: D
Committing a change from a workspace pushes it to the connected Git branch. However, deployment pipelines promote items between workspaces. The Sales_Prod workspace is unaware of the Git commit. To get the new notebook version into production, the team must first go to the Sales_Prod workspace and perform an 'Update' action from source control to pull the changes from the main branch. A deployment pipeline alone does not sync changes from Git.
Question 5
You are processing real-time IoT data using a Fabric Eventstream. The incoming JSON data needs to be enriched with reference data stored in a Delta table in a Lakehouse. The enrichment must happen in near real-time as events flow through the system. Which Fabric component is the most suitable for performing this stateful stream-enrichment operation?
Answer and explanation
Correct answer: B
The Eventstream's event processor provides a no-code/low-code graphical interface to design data transformations on the fly. It includes operations like 'Manage fields', 'Filter', and crucially, 'Expand', which can be used to join the streaming data with reference data from a Lakehouse table. This is the most direct and integrated way to perform such enrichment within the Eventstream itself.
Question 6
Multiple answers
You are tasked with implementing security for a new Fabric Lakehouse. The requirements are as follows:
Data engineers must have full control over the Lakehouse, including the ability to read and write to tables and underlying files.
Business analysts must be able to query all data using the SQL endpoint but must be prevented from accessing the underlying files in OneLake.
A specific group of junior analysts must only see data for the 'North America' region from the Sales table.
Which combination of security mechanisms should you use? (Select TWO)
Answer and explanation
Correct answers: A, C
Workspace roles provide the broad level of access control. 'Admin' gives data engineers full control. 'Viewer' allows business analysts to read content, including querying the SQL endpoint, but prevents them from modifying items or accessing files directly, fulfilling the second requirement.
Row-level security (RLS) is the correct mechanism for filtering rows based on user identity or role. You would create a security predicate (e.g., [Region] = 'North America') and apply it to a role that the junior analysts belong to, fulfilling the third requirement.
Question 7
You are troubleshooting a slow-running PySpark notebook in a Fabric workspace. The notebook reads a large Delta table, performs a series of transformations, and then joins it with a small dimension table. You observe from the Spark UI that one particular stage is taking an exceptionally long time and has a significant amount of data shuffle.
The problematic code snippet is: large_df.join(small_df, 'user_id', 'inner')
Which optimization technique should you apply to mitigate this performance bottleneck?
Answer and explanation
Correct answer: C
The scenario describes a classic large-table-to-small-table join, which is a perfect candidate for a broadcast hash join. By default, Spark might choose a shuffle-based join, causing massive data movement. Using from pyspark.sql.functions import broadcast and rewriting the join as large_df.join(broadcast(small_df), 'user_id', 'inner') instructs Spark to send the small DataFrame to every executor, avoiding the costly shuffle of the large DataFrame and dramatically improving performance.
Question 8
A manufacturing company uses Microsoft Fabric to monitor its production line. An Eventstream ingests sensor data, which is then written to a KQL database for real-time dashboarding. A critical requirement is to detect anomalies where the average temperature of a specific sensor (sensorId = 'SN01-T5') exceeds 150 degrees Celsius over a 5-minute period. You need to write a KQL query to create an alert based on this condition.
Which KQL query correctly implements this logic?
Answer and explanation
Correct answer: B
This query correctly implements the logic. It first filters for the specific sensor. Then, summarize avg(temperature) by bin(timestamp, 5m) calculates the average temperature over 5-minute tumbling windows. Finally, where avg_temperature > 150 filters these aggregated results to find the windows where the anomaly condition is met.
Question 9
True or False: When you use the Mirroring feature in Microsoft Fabric to replicate an Azure SQL Database, the initial setup performs a full snapshot of the source data, and subsequent changes are then replicated in near real-time using change data capture (CDC), without requiring the configuration of a separate data pipeline.
Answer and explanation
Correct answer: A
This statement is true. The Fabric Mirroring feature is designed for seamless, low-latency replication. It automatically handles the initial snapshot of the source database tables and then uses the source's change data capture (CDC) mechanism to replicate ongoing inserts, updates, and deletes to OneLake in Delta format. This provides a continuously updated, queryable copy without the need to build and manage a manual ETL pipeline.
Question 10
Case Study: Global E-Commerce Platform
Company Background GlobalCart is a multinational e-commerce company that operates in North America, Europe, and Asia. They are migrating their analytics platform to Microsoft Fabric to create a unified view of their sales, customer, and inventory data. Their primary goal is to empower regional business units with self-service analytics while maintaining centralized governance and data engineering standards.
Existing Environment Data is currently stored in a mix of sources: an on-premises SQL Server 2019 database for inventory, an Azure SQL Database for sales transactions, and Parquet files in an ADLS Gen2 account for historical clickstream data. The company has a single Fabric F64 capacity located in 'Central US'. The data engineering team is proficient in SQL and PySpark.
Requirements
Architecture: Implement a medallion architecture using a central 'Corporate' workspace for Bronze and Silver layer Lakehouses. Each region (NA, EU, AS) must have its own separate workspace containing a Gold layer Lakehouse with data filtered specifically for that region.
Data Ingestion: Data from the on-premises SQL Server must be ingested nightly with minimal impact on the source system. Sales data from Azure SQL should be replicated in near real-time.
Transformation: Transformations from Bronze to Silver will be complex and require custom Python libraries. Transformations from Silver to the regional Gold layers will involve filtering and simple aggregations.
Security: Data analysts in each region must only be able to access the Gold layer Lakehouse in their respective regional workspace. They should not have any access to the 'Corporate' workspace or other regions' workspaces.
Problem Statement You need to design the data flow and transformation strategy to populate the regional Gold layer Lakehouses from the centralized Silver layer Lakehouse. The solution must be efficient and scalable. Which approach best meets the requirements?
Answer and explanation
Correct answer: B
This centralized approach aligns with the company's goal of maintaining central data engineering standards. A master pipeline in the 'Corporate' workspace can orchestrate the entire Silver-to-Gold process. Using notebooks for the transformation allows for the simple filtering and aggregation logic required. This design provides a single point of management and monitoring for the population of all Gold layers, ensuring consistency and leveraging the central team's expertise. It also avoids granting cross-workspace permissions to regional resources.
