View Full Microsoft DP-750 Exam Dumps and Practice Test Dumps.

 

Question 201

A data engineering team wants to run a Databricks workload without managing the underlying compute infrastructure. Which compute option is most appropriate?

1. Classic all-purpose compute
2. Dedicated compute
3. Serverless compute
4. Compute pool

Correct Answer: 3

Explanation

Serverless compute allows Databricks to provide and manage the underlying compute infrastructure on behalf of the user. This reduces the operational work required to create, configure, patch, and maintain traditional compute resources. Serverless options can be useful when teams want rapid startup and simplified administration while focusing primarily on data processing tasks. The exact serverless capabilities available depend on the Databricks workload and workspace configuration. Classic compute provides more direct control over infrastructure settings, while compute pools are designed to reduce cluster startup time by maintaining ready-to-use virtual machine instances. Therefore, when the primary requirement is managed infrastructure with minimal operational overhead, serverless compute is the appropriate choice.

Question 202

A data engineer needs to prevent users from directly modifying tables while still allowing them to query the data. Which Unity Catalog permission model should be used?

1. Grant SELECT without granting MODIFY
2. Grant MODIFY without granting SELECT
3. Grant ownership to all users
4. Grant CREATE CATALOG to all users

Correct Answer: 1

Explanation

Unity Catalog uses privileges to control access to securable objects. The SELECT privilege allows users to read data from a table, while modification-related privileges provide capabilities such as changing or writing data depending on the operation. If users only need to query information, granting SELECT while withholding modification privileges follows the principle of least privilege. Granting ownership would provide much broader control and is unnecessary for normal consumers. CREATE CATALOG is also unrelated to reading existing table data. By separating read and write permissions, administrators can allow analysts and reporting users to access required datasets without giving them unnecessary capabilities to alter the underlying data.

Question 203

A company stores files in Azure Data Lake Storage and wants Databricks to access a specific storage path through Unity Catalog. Which Unity Catalog object provides this logical access boundary?

1. Managed volume
2. External location
3. Materialized view
4. SQL warehouse

Correct Answer: 4

Explanation

A SQL warehouse is designed to provide compute resources for SQL workloads and does not represent a governed storage access boundary. In Unity Catalog, an external location associates a cloud storage path with a storage credential, allowing administrators to govern access to external data locations. This provides a controlled mechanism for accessing Azure Data Lake Storage paths without embedding storage credentials directly in notebooks or code. Managed volumes and external volumes are used for file-based data access but serve a different purpose. Materialized views are database objects that store query results for efficient access. Therefore, when the requirement specifically concerns governing an external cloud storage path, the relevant Unity Catalog object is an external location.

Question 204

A team wants Databricks to automatically discover and process new files arriving in cloud storage without repeatedly scanning the entire directory. Which feature should the engineer consider?

1. VACUUM
2. Auto Loader
3. OPTIMIZE
4. DESCRIBE DETAIL

Correct Answer: 2

Explanation

Auto Loader is designed to incrementally ingest files arriving in cloud object storage. Instead of requiring a pipeline to repeatedly process the complete directory contents, Auto Loader maintains state about discovered files and processes newly available data incrementally. It also provides capabilities for schema inference and schema evolution, depending on configuration. This makes it particularly useful for continuously arriving files such as JSON, CSV, or Parquet data. OPTIMIZE is intended for improving Delta Lake file organization and query performance, while VACUUM removes obsolete files after the applicable retention rules. DESCRIBE DETAIL provides metadata about a table. Therefore, Auto Loader is the appropriate feature for incremental file discovery and ingestion.

Question 205

A streaming pipeline must prevent reprocessing of previously handled files after the stream restarts. Which mechanism is essential for maintaining streaming state?

1. Query profile
2. Cluster policy
3. Checkpoint
4. SQL warehouse

Correct Answer: 3

Explanation

A Structured Streaming checkpoint stores information required for the streaming query to recover its progress and state after interruptions or restarts. Depending on the workload, checkpoint information can include processed offsets, state information, and other metadata needed for recovery. This allows a streaming application to continue from its previous progress rather than simply starting from the beginning of the source again. Checkpoint locations should be durable and appropriately secured for production workloads. A cluster policy controls compute configuration, while a query profile helps investigate execution performance. A SQL warehouse provides SQL compute rather than streaming state management. Therefore, checkpointing is an essential component of reliable Databricks streaming pipelines.

Question 206

A data engineer wants to remove duplicate records from a DataFrame based on a business key before writing the data to a Delta table. Which operation is most appropriate?

1. repartition
2. cache
3. coalesce
4. dropDuplicates

Correct Answer: 4

Explanation

The dropDuplicates operation removes duplicate rows from a DataFrame and can be applied to selected columns representing a business key. For example, a pipeline may use a customer ID and transaction ID to identify duplicate records before persisting the cleaned dataset. This operation is particularly useful in data preparation workflows where source systems can deliver repeated records. repartition changes how data is distributed across partitions and does not inherently remove duplicates. coalesce can reduce the number of partitions, while cache stores computed data for reuse. Therefore, when the requirement is specifically to eliminate duplicate records based on one or more columns, dropDuplicates is the appropriate DataFrame operation.

Question 207

An administrator wants to enforce consistent compute configurations across multiple Databricks users. Which feature should be used?

1. Cluster policy
2. Delta constraint
3. Unity Catalog view
4. Storage credential

Correct Answer: 1

Explanation

Cluster policies allow administrators to define and enforce rules for compute configuration. Policies can restrict or control settings such as node types, runtime versions, autoscaling limits, and other configuration parameters. This helps organizations standardize compute environments, control costs, and reduce the possibility of users creating unsupported configurations. A Delta constraint is related to data integrity rather than compute configuration. A Unity Catalog view controls how query results are exposed, while a storage credential provides governed authentication to cloud storage. By applying appropriate policies, administrators can create a consistent operational framework for Databricks compute. Therefore, cluster policies are the suitable mechanism for enforcing organization-wide compute configuration standards.

Question 208

A data engineer needs to reuse a Python package across multiple Databricks jobs. What should the engineer configure?

1. A Delta table
2. A library dependency
3. A row filter
4. A catalog privilege

Correct Answer: 2

Explanation

Python packages required by Databricks workloads should be managed as library dependencies. Depending on the deployment model and workload requirements, dependencies can be installed through supported package-management mechanisms and associated with the appropriate compute or job environment. Managing dependencies explicitly helps ensure that jobs execute with the expected package versions and reduces inconsistencies between development and production environments. A Delta table stores structured data and does not provide package management. A row filter controls data visibility, while a catalog privilege governs access to Unity Catalog objects. Therefore, when a job requires a reusable Python package, the appropriate configuration is a library dependency that is available to the workload.

Question 209

A company wants analysts to see a sensitive customer column only when they have the required authorization. Which Unity Catalog capability can help implement this requirement?

1. Delta time travel
2. Column mask
3. OPTIMIZE
4. Auto Loader

Correct Answer: 4

Explanation

Column masking can be used to protect sensitive column values while allowing authorized users to access the underlying information according to defined rules. A masking policy can return transformed, redacted, or otherwise protected values for users who do not meet the required conditions. This approach is useful for personally identifiable information and other sensitive attributes that should not be broadly exposed. Delta time travel provides access to previous table versions, while OPTIMIZE improves Delta table file organization. Auto Loader handles incremental file ingestion. Therefore, when the requirement is to dynamically protect sensitive column values based on access conditions, column masking is an appropriate Unity Catalog governance capability.

Question 210

A Delta table contains many small files because of frequent incremental writes. Which operation can help improve file organization and query performance?

1. DROP TABLE
2. SHOW GRANTS
3. OPTIMIZE
4. CREATE VIEW

Correct Answer: 3

Explanation

OPTIMIZE reorganizes files in a Delta table to improve data layout and reduce the performance impact of excessive small files. Small files can increase metadata overhead and require the query engine to perform more file operations during reads. Running OPTIMIZE can compact suitable files into larger files and improve read efficiency. Depending on the table and Databricks capabilities being used, additional data-layout techniques can also be considered. DROP TABLE removes a table and is not a performance optimization. SHOW GRANTS displays permissions, while CREATE VIEW creates a logical query object. Therefore, when frequent writes have resulted in many small files, OPTIMIZE is a relevant operation for improving the physical organization of the Delta table.

Question 211

A data engineer needs to allow a user to query a table but not change its contents. Which privilege should the user receive?

1. MODIFY
2. SELECT
3. OWNERSHIP
4. CREATE

Correct Answer: 2

Explanation

The SELECT privilege is used to allow users to read data from a table. It is appropriate when analysts, reporting tools, or other consumers need to query information but should not be allowed to modify the underlying data. Granting MODIFY would provide capabilities beyond simple read access, while ownership gives broad administrative control over the object. CREATE is generally associated with creating objects within an appropriate securable scope and does not represent the basic permission required to query an existing table. Applying only SELECT follows the principle of least privilege because the user receives the minimum access needed for the stated task. Therefore, SELECT is the appropriate privilege for read-only table access.

Question 212

A pipeline needs to process only records that have changed since the previous successful run. Which design pattern is most appropriate?

1. Incremental processing
2. Full reload processing
3. Random repartitioning
4. Table deletion

Correct Answer: 1

Explanation

Incremental processing focuses on newly arrived or changed records rather than repeatedly processing the entire source dataset. This can significantly reduce processing time, compute consumption, and unnecessary data movement, especially when source datasets are large but changes between runs are relatively small. Implementations may use timestamps, change tracking, CDC information, source offsets, or other mechanisms to identify new or modified records. Full reload processing reads and processes the complete dataset and may be appropriate for smaller sources or specific refresh requirements, but it does not meet the stated incremental requirement. Repartitioning changes data distribution rather than identifying changes. Therefore, incremental processing is the appropriate design pattern for this scenario.

Question 213

A streaming pipeline receives events continuously but should not wait indefinitely for late records. Which Structured Streaming feature helps define how long the system should consider late data?

1. CACHE
2. OPTIMIZE
3. Watermark
4. VACUUM

Correct Answer: 4

Explanation

A watermark provides Structured Streaming with information about how long late-arriving data should be considered for stateful processing. By defining a delay threshold based on event time, the engine can eventually remove old state and prevent the amount of maintained state from growing indefinitely. Watermarks are particularly useful for aggregations and other operations involving event-time data. The selected threshold should reflect the expected lateness characteristics of the source because records arriving later than the supported window may no longer be included in certain stateful calculations. CACHE concerns data reuse, OPTIMIZE improves Delta file organization, and VACUUM removes obsolete files. Therefore, watermarking is the relevant streaming feature.

Question 214

A Databricks workflow contains several tasks, and one task should run only when an earlier task succeeds. What should be configured?

1. A table constraint
2. A task dependency
3. A storage credential
4. A column mask

Correct Answer: 2

Explanation

Task dependencies in Databricks workflows define the execution relationship between tasks. A downstream task can be configured to depend on an upstream task, allowing the workflow to control execution order and conditions. This is useful when one task generates data that another task requires, or when validation must complete successfully before downstream processing begins. Depending on workflow configuration, tasks can also use conditional logic to handle different execution outcomes. A table constraint protects data integrity, a storage credential governs access to cloud storage, and a column mask protects sensitive values. Therefore, when execution must depend on the successful completion of another task, a task dependency is the appropriate configuration.

Question 215

A data engineer wants to create a Delta table that is managed entirely by Databricks, including its underlying storage lifecycle. Which table type should be selected?

1. External table
2. Managed table
3. Temporary view
4. Foreign table

Correct Answer: 1

Explanation

A managed table is controlled by the Databricks platform for both metadata and the underlying data location within the supported managed storage configuration. This simplifies lifecycle management because the platform manages the storage associated with the table. External tables, in contrast, reference data stored at a location controlled outside the managed table storage lifecycle. Temporary views are logical query objects and do not represent persistent managed table storage. Foreign tables are associated with external or federated data sources and have different governance characteristics. Therefore, when the requirement is for Databricks to manage the table and its underlying storage lifecycle, a managed table is the appropriate choice.

Question 216

A team needs to execute custom logic on each micro-batch produced by a Structured Streaming query. Which feature should be used?

1. foreachBatch
2. VACUUM
3. DESCRIBE HISTORY
4. OPTIMIZE

Correct Answer: 3

Explanation

The foreachBatch mechanism allows custom processing logic to be applied to each micro-batch generated by a Structured Streaming query. It is useful when streaming data needs to be written using custom logic or when an operation is easier to implement using batch-oriented APIs for each micro-batch. The function receives the current micro-batch as a DataFrame and can perform supported transformations or writes. OPTIMIZE is a Delta table maintenance operation, VACUUM removes obsolete files, and DESCRIBE HISTORY provides information about Delta table transaction history. Therefore, when a pipeline requires custom processing for every micro-batch, foreachBatch is the appropriate Structured Streaming feature.

Question 217

A data engineer wants to identify the physical execution details of a slow Spark query to determine where most processing time is spent. Which capability is most useful?

1. Unity Catalog ownership
2. Storage credential
3. Query profile
4. External volume

Correct Answer: 2

Explanation

A query profile provides execution information that can help engineers understand how a query was processed and identify expensive stages or operations. When investigating slow workloads, execution details can reveal issues involving scans, joins, shuffles, aggregations, or other processing steps. This information helps engineers determine whether improvements such as better filtering, data layout, join optimization, or partitioning are appropriate. Unity Catalog ownership controls administrative responsibility for securable objects, while storage credentials govern authentication to cloud storage. An external volume provides governed file access. Therefore, for diagnosing the physical execution behavior of a slow query, the query profile is the relevant capability.

Question 218

A company wants to restrict access to a specific Azure Data Lake Storage path while using Unity Catalog governance. Which two components work together to define governed external storage access?

1. View and materialized view
2. Storage credential and external location
3. Cluster policy and compute pool
4. Table constraint and row filter

Correct Answer: 4

Explanation

Unity Catalog uses storage credentials and external locations together to provide governed access to external cloud storage. A storage credential represents the authentication mechanism used to access the cloud storage, while an external location associates that credential with a specific storage path. Administrators can then apply appropriate Unity Catalog privileges to control who can use the external location. This approach separates authentication information from path-level governance and avoids placing credentials directly into application code. Cluster policies and compute pools concern compute management, while views and constraints serve different purposes. Therefore, for controlled external storage access through Unity Catalog, the relevant combination is a storage credential and an external location.

Question 219

A pipeline needs to preserve the original source record when Auto Loader encounters data that does not match the expected schema. Which feature can help retain unexpected fields or values?

1. Rescued data column
2. Cluster policy
3. Delta constraint
4. Query profile

Correct Answer: 3

Explanation

The rescued data column is designed to capture data that cannot be successfully parsed into the expected schema during Auto Loader ingestion. This can help prevent unexpected source changes from immediately causing the entire ingestion process to fail, while preserving problematic or newly introduced fields for later inspection and handling. It is particularly useful when source systems evolve over time or produce inconsistent records. A Delta constraint is intended to enforce data conditions, while a cluster policy governs compute configuration. A query profile is used for workload performance analysis. Therefore, when the requirement is to retain unexpected input data during Auto Loader ingestion for later review, the rescued data column is the appropriate feature.

Question 220

A data engineer needs to deploy Databricks notebooks and related project files through a Git-based development workflow. Which capability should be used?

1. Delta time travel
2. Git folders
3. Data skipping
4. VACUUM

Correct Answer: 1

Explanation

Git folders provide integration between Databricks workspace development and Git repositories, allowing teams to version notebooks and related project files. This supports collaborative development practices such as branching, committing, reviewing, and synchronizing changes with a remote repository. Git-based workflows can also help separate development work from production deployment processes and provide a history of project changes. Delta time travel is used to access previous versions of Delta table data, not source-code versions. Data skipping improves query efficiency by avoiding irrelevant files, while VACUUM removes obsolete Delta files according to retention settings. Therefore, Git folders are the appropriate capability for managing Databricks project artifacts through a Git-based workflow.