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

 

Question 341

A data engineer needs to combine records from a source dataset with an existing Delta table, updating matching records and inserting new records. Which operation should be used?

1. OPTIMIZE
2. VACUUM
3. MERGE
4. DESCRIBE DETAIL

Correct Answer: 3

Explanation

The MERGE operation is designed to synchronize records between a source dataset and a target Delta table. It can update existing records when a matching condition is satisfied and insert new records when no matching target record exists. This makes MERGE especially useful for incremental data processing and upsert scenarios. It can also support conditional update and insert logic depending on the pipeline requirements. OPTIMIZE focuses on file organization and performance, VACUUM removes eligible obsolete files, and DESCRIBE DETAIL returns table metadata. Therefore, when records need to be updated if they already exist and inserted when they are new, MERGE is the appropriate Delta operation.

Question 342

A data engineer wants to remove duplicate records from a DataFrame based on one or more columns. Which DataFrame operation should be considered?

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

Correct Answer: 1

Explanation

The dropDuplicates operation removes duplicate rows from a DataFrame based on the complete row or on specified columns. It is useful when source data contains repeated records that should not be propagated into downstream tables. For example, a pipeline can use dropDuplicates with a business key to retain only one occurrence of records having the same key. repartition changes the number or distribution of partitions, cache stores computed data for reuse, and coalesce can reduce the number of partitions. These operations do not directly remove duplicate records. Therefore, dropDuplicates is the appropriate DataFrame operation when duplicate input records need to be eliminated.

Question 343

A data engineer wants to reduce the number of partitions in a DataFrame without causing a full shuffle when possible. Which operation is most appropriate?

1. MERGE
2. coalesce
3. VACUUM
4. DESCRIBE HISTORY

Correct Answer: 2

Explanation

The coalesce operation can reduce the number of partitions in a DataFrame while avoiding a full shuffle in common scenarios. This can be useful when a previous transformation has produced more partitions than necessary and the data needs to be written with fewer output partitions. Because coalesce generally avoids redistributing all records across the cluster, it can be more efficient than repartition when only a reduction in partition count is required. MERGE handles Delta table synchronization, VACUUM manages obsolete files, and DESCRIBE HISTORY displays table transaction history. Therefore, coalesce is the suitable operation when reducing partitions efficiently is the primary requirement.

Question 344

A Delta table must retain historical versions so that engineers can investigate previous changes. Which command provides information about the table’s transaction history?

1. SHOW TABLES
2. DESCRIBE HISTORY
3. CREATE VOLUME
4. OPTIMIZE

Correct Answer: 2

Explanation

DESCRIBE HISTORY provides information about transactions performed against a Delta table. The history can help engineers investigate operations such as writes, updates, merges, deletes, and other table changes. This information is useful for auditing, troubleshooting, and understanding how a table reached its current state. SHOW TABLES lists available tables but does not provide transaction history. CREATE VOLUME creates a Unity Catalog volume, while OPTIMIZE improves Delta file organization. Therefore, when an engineer needs to investigate historical table operations and changes, DESCRIBE HISTORY is the appropriate command to use.

Question 345

A streaming pipeline receives events containing an event-time column. The engineer needs to handle events that arrive later than expected while limiting the amount of retained streaming state. Which approach is appropriate?

1. Use a watermark
2. Use a cluster policy
3. Use an external location
4. Use a storage credential

Correct Answer: 1

Explanation

A watermark allows Structured Streaming to track progress in event time and define how much lateness the application is willing to handle for certain stateful operations. This helps balance correctness for late-arriving events against the amount of state retained by the streaming application. Without appropriate state management, long-running streams can accumulate significant state and consume resources. Cluster policies govern compute configuration, external locations govern cloud storage paths, and storage credentials provide authentication to cloud storage. These features do not control event-time state retention. Therefore, when late events and streaming state need to be managed together, watermarking is the appropriate approach.

Question 346

A data engineer needs to divide a DataFrame into a specific number of partitions and redistribute the records across those partitions. Which operation should be used?

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

Correct Answer: 3

Explanation

The repartition operation changes the number of partitions and redistributes records across the resulting partitions. Because it generally involves a shuffle, repartition can be useful when the data needs to be evenly redistributed or partitioned according to specific columns. This can support parallel processing and may improve performance for certain workloads when partitioning is designed appropriately. cache stores computed results for reuse, coalesce reduces partitions with less redistribution when possible, and dropDuplicates removes repeated records. Therefore, when the requirement is to redistribute data into a specified partition structure, repartition is the appropriate DataFrame operation.

Question 347

A team wants users to query a logical representation of data without giving them direct access to the underlying table structure. Which object can provide this abstraction?

1. View
2. Storage credential
3. Cluster policy
4. Volume

Correct Answer: 1

Explanation

A view provides a logical representation of data based on a query. Views can abstract underlying tables and allow users to access selected columns, filtered rows, or transformed data without directly querying the underlying objects in the same way. In a governed environment, permissions can also be applied to control who can access the view and its underlying data. Storage credentials manage authentication to cloud storage, cluster policies control compute configurations, and volumes provide governed access to files. Therefore, when users need a logical SQL representation that abstracts underlying data structures, a view is an appropriate object.

Question 348

A data engineer wants to store non-tabular files such as PDFs, images, and machine-learning artifacts in a governed Unity Catalog environment. Which object is appropriate?

1. Schema
2. Volume
3. View
4. SQL warehouse

Correct Answer: 2

Explanation

Unity Catalog volumes provide a governed location for storing and accessing files that are not necessarily represented as relational tables. Examples include PDFs, images, configuration files, machine-learning artifacts, and other unstructured or semi-structured files. Volumes can be managed through Unity Catalog permissions and integrated into data workflows. A schema organizes database objects such as tables and views, a view represents query results logically, and a SQL warehouse provides SQL compute. Therefore, when a team needs governed storage for non-tabular files within Unity Catalog, a volume is the appropriate object.

Question 349

A data engineer wants to identify the permissions granted on a Unity Catalog object. Which SQL command can provide this information?

1. SHOW GRANTS
2. VACUUM
3. OPTIMIZE
4. DESCRIBE HISTORY

Correct Answer: 1

Explanation

SHOW GRANTS can be used to inspect permissions associated with supported Unity Catalog securable objects. It is useful for administrators and data engineers troubleshooting access problems or auditing which users and groups have privileges on an object. Understanding granted permissions is an important part of implementing governed data access. VACUUM manages obsolete Delta files, OPTIMIZE improves file layout, and DESCRIBE HISTORY reports Delta table transaction history. None of these commands is intended to list object privileges. Therefore, when the requirement is to inspect permissions granted on a governed object, SHOW GRANTS is the appropriate command.

Question 350

A pipeline needs to process only new records from a source system during each scheduled execution rather than reload the entire dataset. Which approach is most appropriate?

1. Full refresh
2. Incremental processing
3. Manual table recreation
4. Permanent caching

Correct Answer: 2

Explanation

Incremental processing focuses each pipeline execution on newly arrived or changed data instead of repeatedly processing the entire source dataset. This can significantly reduce processing time, compute consumption, and unnecessary data movement for large workloads. Incremental pipelines commonly use timestamps, change tracking, source-system identifiers, checkpoints, or merge logic to determine what needs to be processed. A full refresh intentionally reloads the complete dataset, manual table recreation is not an efficient incremental strategy, and permanent caching does not determine which records are new. Therefore, when only new or changed records should be processed during each run, incremental processing is the appropriate approach.

Question 351

A Delta table contains obsolete files that are no longer required according to the applicable retention rules. Which command is designed to remove those files?

1. MERGE
2. VACUUM
3. REPARTITION
4. SHOW GRANTS

Correct Answer: 2

Explanation

VACUUM is used to remove obsolete data files from a Delta table after they are no longer needed according to the configured retention requirements. Removing unnecessary files can help control storage consumption, although retention settings must be considered carefully because historical versions and time-travel access may depend on those files. MERGE synchronizes records, REPARTITION redistributes DataFrame data, and SHOW GRANTS displays permissions. These operations do not perform Delta file cleanup. Therefore, when obsolete Delta files need to be removed in accordance with retention policies, VACUUM is the appropriate command.

Question 352

A data engineer wants a pipeline to continue processing from its previous position after a Structured Streaming application restarts. Which feature should be configured?

1. Checkpoint location
2. Cluster policy
3. Row filter
4. Materialized view

Correct Answer: 1

Explanation

A checkpoint location stores the progress and state information required for reliable recovery of a Structured Streaming application. When the application restarts and uses the appropriate checkpoint information, the streaming engine can recover its previous progress and continue processing according to the stored state. Checkpoints are therefore an important component of fault-tolerant streaming pipelines. Cluster policies control compute settings, row filters restrict visible records, and materialized views provide persisted query results. These features do not store streaming execution progress. Therefore, when a streaming application needs to recover from its previous position after a restart, a checkpoint location should be configured.

Question 353

A team wants to organize Unity Catalog objects according to business domains while maintaining a structured hierarchy. Which object is commonly used to group tables and views within a catalog?

1. Schema
2. Volume
3. Storage credential
4. Cluster policy

Correct Answer: 1

Explanation

A schema provides a logical namespace for organizing objects such as tables, views, and related Unity Catalog resources within a catalog. Organizations can use schemas to separate objects according to departments, applications, environments, or business domains. This hierarchical organization makes data discovery and permission management easier. A volume is designed for file-based data, a storage credential defines authentication for external storage, and a cluster policy controls compute configuration. Therefore, when tables and views need to be organized within a catalog according to logical domains, schemas are an appropriate organizational layer.

Question 354

A query frequently filters a large Delta table using a column that is also useful for organizing the physical data layout. Which optimization approach can improve query efficiency by organizing related data together?

1. Liquid clustering
2. Secret rotation
3. User provisioning
4. Job cancellation

Correct Answer: 1

Explanation

Liquid clustering is a Delta table optimization capability that can organize data according to selected clustering keys while allowing the layout to evolve as workloads and data distributions change. It can improve query performance when filtering or accessing data based on appropriate columns because related records can be organized more efficiently for data skipping and file access. Secret rotation handles credential management, user provisioning manages identities and access, and job cancellation stops running workloads. These features do not optimize physical table layout. Therefore, when a Delta table needs an adaptive data layout that supports common query patterns, liquid clustering can be considered.

Question 355

A data engineer needs to preserve records that fail validation so they can be reviewed separately instead of stopping the entire ingestion pipeline. Which pattern is appropriate?

1. Quarantine invalid records
2. Delete the entire source
3. Disable validation
4. Drop the target table

Correct Answer: 1

Explanation

A quarantine pattern separates records that fail validation from records that successfully pass data-quality checks. Instead of allowing invalid records to enter the trusted dataset or causing the entire pipeline to fail, the pipeline can write rejected records to a designated quarantine location. Data engineers can then inspect, correct, and potentially reprocess those records later. This approach supports resilient data pipelines while preserving visibility into data-quality problems. Deleting the source or target does not solve the validation problem, and disabling validation can allow poor-quality data into downstream systems. Therefore, quarantining invalid records is an appropriate pattern for handling data-quality failures without unnecessarily stopping successful processing.

Question 356

A workflow task should execute only after another task successfully completes. Which workflow configuration should be used?

1. Task dependency
2. Table constraint
3. Data mask
4. Storage credential

Correct Answer: 1

Explanation

A task dependency defines the relationship between workflow tasks and controls when a downstream task becomes eligible to run. If a task must execute only after another task successfully completes, the workflow can establish a dependency between them. This allows complex pipelines to represent processing order and ensures that downstream operations do not begin before required upstream work is finished. Table constraints enforce data-quality rules, data masks protect sensitive values, and storage credentials provide authentication for external storage. None of these features controls task execution order. Therefore, when one workflow task must wait for another task to complete successfully, a task dependency should be configured.

Question 357

A data engineer wants to improve performance when repeatedly accessing the same intermediate DataFrame during a Spark workload. Which operation can be used to retain computed data for reuse?

1. cache
2. VACUUM
3. MERGE
4. DROP SCHEMA

Correct Answer: 1

Explanation

The cache operation can retain the computed representation of a DataFrame so that subsequent actions may reuse it rather than recomputing the same transformations from the beginning. This can improve performance when an intermediate dataset is accessed multiple times during a workload. However, caching should be used thoughtfully because cached data consumes cluster resources and may not benefit datasets that are accessed only once. VACUUM manages obsolete Delta files, MERGE synchronizes Delta records, and DROP SCHEMA removes a schema. Therefore, when an intermediate DataFrame is repeatedly reused and recomputation would be expensive, caching can be an appropriate performance technique.

Question 358

A data engineer wants to monitor whether a scheduled production job is succeeding, failing, or taking longer than expected. Which capability should be used?

1. Job monitoring
2. Column masking
3. External location
4. Schema evolution

Correct Answer: 1

Explanation

Job monitoring provides operational visibility into workflow execution. Engineers can inspect job runs, task states, failures, durations, and other execution information to identify problems and understand production behavior. Monitoring is important for detecting failed tasks, unexpected execution times, repeated retries, and other operational issues. Column masking protects sensitive data, external locations govern access to cloud storage paths, and schema evolution manages changing input structures. These features do not provide operational visibility into workflow executions. Therefore, when a production team needs to determine whether scheduled jobs are succeeding or failing and investigate execution behavior, job monitoring is the appropriate capability.

Question 359

A data engineer needs to join a continuously arriving streaming dataset with a static reference table containing customer information. Which approach can support this requirement?

1. Stream-static join
2. Two unrelated batch jobs
3. VACUUM operation
4. Cluster policy

Correct Answer: 1

Explanation

A stream-static join allows a streaming DataFrame to be combined with a static DataFrame. This pattern is useful when continuously arriving events need enrichment with relatively stable reference information, such as customer attributes, product details, or location mappings. The static dataset can provide descriptive information while the streaming dataset supplies the continuously arriving records. VACUUM is used for Delta file cleanup and cluster policies govern compute configuration. Running unrelated batch jobs does not directly provide the required streaming enrichment operation. Therefore, when a streaming dataset needs to be enriched using a static reference dataset, a stream-static join is an appropriate approach.

Question 360

A data engineer wants to inspect the schema and column definitions of a table before developing a transformation. Which command is useful for this purpose?

1. DESCRIBE TABLE
2. VACUUM
3. OPTIMIZE
4. MERGE

Correct Answer: 1

Explanation

DESCRIBE TABLE is useful for inspecting information about a table’s structure, including its columns and associated metadata. Understanding the schema before writing transformations helps engineers reference the correct column names and data types and avoid errors during development. It can also be useful when investigating unfamiliar datasets or validating that a table matches expected definitions. VACUUM removes eligible obsolete files, OPTIMIZE improves Delta file organization, and MERGE performs source-to-target synchronization. These commands do not primarily provide the table’s schema definition. Therefore, when a data engineer needs to inspect table columns and structure before creating a transformation, DESCRIBE TABLE is appropriate.