DP-600 — frequently asked questions

According to current exam standards and documentation, Fabric natively supports only Azure Repos for Git integration within the same tenant.

No. OneDrive is a file storage service and does not provide Git-based branching or version control features required by the requirement.

Comparing factors describes 'what' is happening across segments. Diagnostic analytics requires actively investigating 'why' those differences exist through causal analysis.

Descriptive summarizes historical data to show trends and distributions. Diagnostic analyzes that data to identify root causes and reasons behind those trends.

ISEMPTY checks for the existence of any row without counting them, which avoids aggregation overhead.

Wrap the function in NOT, e.g., NOT ISEMPTY(...), to keep rows where the table is not empty.

Yes, it is valid. It evaluates the table returned by CALCULATETABLE and returns TRUE if any rows exist, acting similarly to an EXISTS clause in SQL.

COUNTROWS must iterate through all matching rows to calculate the total number, while ISEMPTY can stop processing as soon as it finds the first row, saving computational resources.

No, unless a table or view has been explicitly created in the Lakehouse catalog pointing to the shortcut. Otherwise, use the Files/ path.

The Files/ prefix identifies the OneLake file system root, allowing Spark to resolve the shortcut location as a directory structure.

It only computes count, min, and max for strings. Mean and stddev are mathematically undefined for non-numeric data.

Use summary() with explicit aggregation functions or filter columns by type before applying describe().

No, the SQL analytics endpoint is strictly read-only. Write operations must be performed using Spark notebooks or dataflows.

External tables created via Spark are not registered in the SQL catalog, so they are invisible to the SQL endpoint's T-SQL interface.

Because the user object must be synced to Microsoft Entra ID. If it's not in the sync list, it doesn't exist in the cloud identity store.

USERPRINCIPALNAME() is often safer for clarity, but both require the user to be successfully authenticated and present in the directory.

Table.Max returns the single row with the highest value in one specified column. It cannot generate a summary of max values for all numeric columns at once.

It returns a table with statistics like Min, Max, Average, Standard Deviation, and Count for every column in the source table.

T-SQL runs in the SQL Pool engine, whereas Fabric Notebooks run in the Spark pool. The PREDICT function in this context refers to Spark-based ML inference, not SQL Server's predictive analytics.

Yes, SparkR is listed as a supported language alongside PySpark and Spark SQL. However, PySpark is generally preferred due to better ecosystem integration with libraries like MLflow and scikit-learn.

No. Moving an item changes its path, but it remains part of the workspace content. The deployment pipeline deploys all items, preserving their new folder paths.

Pipeline1 was never moved into a folder in Workspace1. Therefore, it stays at the root level in the target workspace after deployment.

In Python visuals, the primary method to disable grouping is ensuring row uniqueness via an index field, as the Summarize By property is less reliable for this specific visual type.

Option A refers to Python coding syntax (iloc), while Option C refers to adding a column to the underlying dataset to force distinctness before the script runs.

Tenant settings enable the XMLA endpoint globally. Since the endpoint is already present, the specific read-write toggle is managed at the Capacity level.

No. XMLA read-write access is a property of the Capacity hosting the dataset, not individual datasets or workspaces.

The 'Users can edit data model in the Power BI service' setting explicitly excludes Direct Lake models and XMLA-based editing according to Microsoft documentation.

Tenant settings (A) enable the feature globally, while Capacity settings (D) define the permission level (Read/Write). Both are required for external tools to publish changes.

PBIX is a binary format. It cannot be read or written by TMDL scripts or editors like VS Code. PBIP uses plain text files for metadata.

PBIP is a project format for development and collaboration. PBIT is a static template file used to generate new reports with predefined structures.

Admin access to the pipeline allows users to view pipeline settings and initiate deployment steps. Without this, even Contributor access to workspaces won't allow triggering the pipeline flow.

While Member access includes Contributor capabilities, the exam options specify Contributor. Contributor is the precise role needed for deploying items. Using Member might imply additional management rights not strictly required for just deploying items.

Caching stores results for the first page load but does not consistently optimize queries for new visuals or pages as effectively as aggregations which reduce the underlying data volume.

User-defined aggregations require manual creation of summary tables, while Automatic aggregations are created by the service based on query patterns without manual intervention.

Traditionally no, but in Microsoft Fabric with OneLake, yes. OneLake supports query folding for CSVs, enabling efficient filtering at the source.

Power BI pulls all raw data into the service memory to apply transformations locally, which can quickly exhaust available resources and cause refresh failures.

Yes, but not natively against parquet files. It supports RLS by falling back to DirectQuery mode to query the SQL endpoint.

Connecting to tables/views in the SQL analytics endpoint that enforce RLS or contain T-SQL views triggers a fallback to DirectQuery.

Analyze in Excel connects to the model for PivotTable exploration and does not expose column-level memory or storage segment details.

No, it reports memory grants for queries in progress, not the column segments stored in memory.

Calculated columns are stored in the model and consume memory, so they increase model size and refresh time instead of reducing them.

Separating the datetime column lowers the number of distinct values in each resulting column and improves VertiPaq compression, which reduces memory usage.

The Fabric SQL analytics endpoint exposes only Delta tables. A Parquet shortcut in Tables is not materialized as a Delta table, so it remains a file object and cannot be queried through the SQL endpoint.

Files-section shortcuts expose raw objects for Spark or notebook reads but do not create table metadata. The SQL analytics endpoint therefore cannot see the S3 objects as queryable tables.

Optimize only applies to existing Delta tables to compact files and apply V-Order; it cannot read CSV files or change the table format.

Yes. Tables created via Load to Tables use the Delta Lake format with V-Order optimization enabled by default.

Append only inserts rows into the existing unpartitioned Table1, so the Enable partition option stays hidden. Fabric lets you define partition columns only when Overwrite re-creates the Delta table with that partition scheme.

No. Source partition discovery reads partition folders from the source files, while destination partitioning for Table1 is configured on the Destination tab under Advanced, after Table action is set to Overwrite.

It inserts a productline1 folder that Lakehouse1 never has; the shortcut is created directly under Tables, so the correct path would be Tables/ResearchProduct, not with an extra level.

No. external_table is KQL syntax for querying external tables in Eventhouse/KQL databases, not PySpark notebook code against Lakehouse1 tables.

Lakehouse V-Order depends on the engine that writes the table, and compaction needs you to run OPTIMIZE ... VORDER yourself, so the automatic guarantee the question demands is not assured.

Yes. Dataflow Gen2 supports a Warehouse destination with the Append update method, so the folder on the local computer can be loaded and appended without pipelines.

Hourly recurrence applies to all days continuously. It lacks the UI element to filter execution to only specific weekdays like Monday and Friday.

Yes, when combined with the 'Repeat every' setting. You select Weekly, pick Mon/Fri, and set the interval to 4 hours.

sys.dm_pdw_exec_requests is specific to Azure Synapse Analytics (SQL DW). Fabric warehouses use standard SQL Server DMVs like sys.dm_exec_requests.

Microsoft's official documentation for Fabric Data Warehouse monitoring lists all supported DMVs, including sys.dm_exec_requests for query diagnosis.

In a Fabric Lakehouse, the SQL analytics endpoint does not support creating or executing stored procedures; that feature is exclusive to the Warehouse.

A view with a recursive CTE and GETDATE() dynamically generates the date range each time it is queried, so it always includes the current year without manual refresh.

Direct Lake semantic models live in a Fabric workspace and Desktop cannot open them, so the calculation group must be added through the XMLA endpoint with a modeling tool like Tabular Editor.

The service can author calculation groups for Direct Lake models via model explorer, but the DP-600 scenario expects the XMLA modeling tool Tabular Editor to implement them.

Copy data moves files with connector mappings but has no Power Query M engine, so it fails the stated requirement that the activity support M expressions.

Yes. Dataflow Gen2 supports a Lakehouse destination, so the same activity that evaluates your M transformations can load the CSV data into Lakehouse1.

Static roles show all data allowed by the role. Dynamic roles require a specific user identity to apply filters correctly, so testing as a user is necessary.

Yes, administrators can use 'Test as role' in the Power BI Service portal to simulate access for specific users assigned to roles.

Workspace identities manage access permissions and role assignments, not data classification or security policies like DLP.

No, in Microsoft 365 and Fabric, DLP policies are typically triggered and enforced through the application of sensitivity labels to the content.

Show only renders text output. It does not create the interactive widget required for switching to the Chart tab.

Yes, the display function supports both Spark DataFrames and Resilient Distributed Datasets (RDDs) in Fabric notebooks.

No. Default sharing grants SQL endpoint access. Report building requires access to the linked semantic model/dataset.

No. The default permission is 'Read'. Write/Edit permissions must be explicitly assigned.

In Fabric Data Factory, the Stored Procedure activity executes the code but does not support returning output parameters or result sets to the pipeline.

If the Lookup activity were an option, it would be preferred for simple value retrieval. However, for general script execution with outputs, Script is the correct choice among the provided options.

Direct Lake optimizes read/query performance. For write operations via XMLA, the Large Model format is the specific feature designed to enhance write throughput and compression.

No. While it reduces storage, it does not engage the specialized engine optimizations provided by the Large semantic model format for write acceleration.

'Other' represents time spent on non-data-preparation tasks, such as waiting for other visuals to finish rendering, background processing, or UI preparation.

The DAX query duration is only 27 ms, which is negligible compared to the 1047 ms in 'Other'. Optimizing a fast query won't solve a bottleneck caused by visual contention.

It allows fallback to DirectQuery mode if data cannot be efficiently loaded into memory, which violates the requirement for pure Direct Lake.

No, Direct Lake Behavior is a model-level property. Partition settings control storage and refresh, not the global query fallback logic.

Descriptive analytics summarizes what happened (e.g., histograms), while diagnostic analytics investigates why it happened (e.g., correlation analysis).

No, a histogram describes the current or past distribution of data. Predictive analytics would use this data to forecast future trends.

ProductID is only part of Product's composite primary key, so it repeats across companies and would not uniquely match each row, producing incorrect order-item analysis.

Denormalizing company attributes into both tables preserves the data while letting you remove the Company table, keeping a simple star schema for analysis by date, product, and customer.

In Microsoft Fabric Copy activities, partition column configuration is only exposed when the Table action is set to Overwrite.

Partition discovery helps identify source folder structures, but it is not the primary step for defining destination partition columns.

No, GitHub is now supported as of September 2024. However, this specific exam question relies on older documentation where only Azure Repos was available.

OneDrive is a file hosting service. It does not provide the structured Git-based version control, branching, and CI/CD pipeline integration required by Fabric workspaces.