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

 

Question 281

A data engineer needs to ingest files from Azure Data Lake Storage as they arrive and automatically track which files have already been processed. Which feature is designed for this requirement?

1. Auto Loader
2. VACUUM
3. Materialized view
4. Cluster policy

Correct Answer: 1

Explanation

Auto Loader is designed for incremental file ingestion from cloud object storage. It keeps track of files that have already been discovered and processed, allowing pipelines to focus on newly arriving data instead of repeatedly scanning and processing the complete source directory. This makes Auto Loader suitable for scalable ingestion workloads where files arrive continuously or in frequent batches. VACUUM is used to remove obsolete Delta files, while materialized views store query-derived results and cluster policies control compute configuration. Auto Loader can also provide schema inference and schema evolution capabilities depending on the configured options. Therefore, Auto Loader is the appropriate feature for incremental file ingestion and file tracking.

Question 282

A data engineer needs to allow a group to create tables within a specific Unity Catalog schema. Which privilege is most directly related to this requirement?

1. SELECT
2. MODIFY
3. CREATE TABLE
4. READ VOLUME

Correct Answer: 3

Explanation

The CREATE TABLE privilege allows an authorized user or group to create tables within an appropriate schema when the required parent-level permissions are also satisfied. Granting only SELECT allows users to read data and does not provide table-creation capability. MODIFY is related to changing data in supported objects but does not by itself represent the required permission to create a new table. READ VOLUME applies to reading files from governed volumes. Access should be granted at the narrowest appropriate scope to follow the principle of least privilege. Therefore, when a group specifically needs to create tables in a schema, the CREATE TABLE privilege is the relevant permission.

Question 283

A streaming pipeline receives events with timestamps that can arrive out of order. Which concept should the engineer use to process records according to when the events actually occurred?

1. Processing-time partitioning
2. Event-time processing
3. Cluster autoscaling
4. File compaction

Correct Answer: 2

Explanation

Event-time processing uses the timestamp associated with an event rather than relying only on the time when the processing system receives the record. This is important when events can arrive late or out of order because network delays, source-system behavior, and distributed processing can cause records to reach the pipeline after newer events. Event-time processing is commonly combined with watermarking for stateful streaming operations so that the system can handle expected late-arriving records while eventually cleaning up old state. Cluster autoscaling manages compute resources, and file compaction improves storage organization. Therefore, when the business meaning depends on when an event actually occurred, event-time processing is the appropriate concept.

Question 284

A company wants to prevent users from creating compute resources that exceed a defined maximum autoscaling range. Which feature should administrators configure?

1. View
2. Cluster policy
3. Delta table
4. External volume

Correct Answer: 2

Explanation

Cluster policies allow administrators to control compute configuration and establish limits on settings such as minimum and maximum worker counts. This can help organizations manage costs and ensure that users create resources within approved operational boundaries. A policy can also standardize runtime versions, node types, and other supported settings. Views provide logical access to data, Delta tables store structured data, and external volumes provide governed access to external files. None of these objects directly restricts compute sizing. Therefore, when an administrator needs to prevent users from exceeding a defined autoscaling range, a cluster policy is the appropriate mechanism for enforcing the limit.

Question 285

A data engineer wants to combine streaming data with a relatively static reference dataset containing customer information. Which type of operation can support this design?

1. Stream-static join
2. VACUUM operation
3. Cluster policy
4. Storage credential

Correct Answer: 1

Explanation

A stream-static join combines a continuously arriving streaming dataset with a relatively static or slowly changing dataset. For example, an event stream can be joined with a customer reference table to enrich each incoming event with customer attributes. This pattern is useful when streaming records require additional context that already exists in a batch-oriented dataset. VACUUM removes obsolete Delta files, cluster policies manage compute configuration, and storage credentials provide authentication for external cloud storage. These features do not perform dataset joins. Therefore, when a streaming workload needs to enrich incoming records with information from a static reference dataset, a stream-static join is an appropriate design pattern.

Question 286

A data engineer wants to inspect the permissions granted on a Unity Catalog schema before troubleshooting an access issue. Which command is useful?

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

Correct Answer: 1

Explanation

SHOW GRANTS can be used to inspect permissions associated with Unity Catalog securable objects, including schemas. Reviewing grants is an important troubleshooting step when a user reports that they cannot access a table, view, volume, or other governed object. Administrators can determine whether the required privileges have been assigned directly or through applicable group memberships and inheritance. OPTIMIZE and VACUUM are Delta maintenance operations, while DESCRIBE HISTORY provides transaction history for Delta tables. These commands do not primarily display access permissions. Therefore, when the objective is to investigate which privileges are assigned to a Unity Catalog schema, SHOW GRANTS is the appropriate command.

Question 287

A pipeline must preserve raw source data before applying cleansing and business transformations. Which medallion layer is normally used for this purpose?

1. Gold
2. Silver
3. Bronze
4. Presentation

Correct Answer: 3

Explanation

The Bronze layer is generally used to preserve raw or minimally transformed source data. It provides an initial landing layer where information can be retained before quality checks, standardization, enrichment, and business transformations are applied. The Silver layer typically contains cleaned and validated data, while the Gold layer contains curated datasets designed for business reporting and analytics. Keeping raw information in Bronze can help with traceability, replay, troubleshooting, and downstream reprocessing when transformation logic changes. Presentation is not one of the standard names in the traditional three-layer medallion architecture. Therefore, when the requirement is to preserve source data before transformation, the Bronze layer is appropriate.

Question 288

A data engineer wants to identify how much data a query reads compared with how much data it ultimately returns. Which capability can help analyze query execution?

1. Query profile
2. Storage credential
3. Volume privilege
4. Job schedule

Correct Answer: 1

Explanation

A query profile provides execution information that can help engineers understand how a query processes data. Reviewing execution details can help identify expensive scans, joins, aggregations, shuffles, and other operations that contribute to resource consumption. It can also help engineers investigate whether filtering and data-layout techniques are effectively reducing unnecessary data processing. Storage credentials manage authentication to external storage, volume privileges control file access, and job schedules determine when workflows execute. These features do not provide detailed information about query execution behavior. Therefore, when the goal is to investigate the amount of data processed and the operations responsible for query cost, the query profile is a useful diagnostic capability.

Question 289

A team needs to maintain a table containing historical versions so that analysts can query an earlier state of the data. Which Delta Lake capability supports this requirement?

1. Time travel
2. Cluster policy
3. Autoscaling
4. Auto Loader

Correct Answer: 1

Explanation

Delta Lake time travel allows users to access previous versions of a Delta table according to the available transaction history and retention configuration. It can be useful for auditing, troubleshooting, reproducing previous analytical results, and investigating how data changed over time. The ability to query historical versions depends on retaining the required transaction and data files, so maintenance operations and retention settings should be considered carefully. Cluster policies and autoscaling manage compute resources, while Auto Loader handles incremental file ingestion. These features do not provide historical table-version access. Therefore, when analysts need to inspect an earlier state of a Delta table, time travel is the appropriate capability.

Question 290

A data engineering team wants a workflow to continue running independent tasks even when another unrelated task fails. Which workflow design should be considered?

1. Independent task dependencies
2. Single mandatory dependency chain
3. Shared storage credential
4. Table partitioning

Correct Answer: 1

Explanation

Workflow tasks should be connected through dependencies only when there is a logical relationship between their execution requirements. If two tasks are independent, they can be configured without making one depend on the successful completion of the other. This allows an unrelated failure to avoid unnecessarily blocking other work. A single mandatory dependency chain would force downstream tasks to wait for upstream tasks even when no data dependency exists. Storage credentials govern external storage authentication, while table partitioning affects data organization. Therefore, designing independent tasks with only the necessary dependencies can improve workflow resilience and allow unrelated processing to continue when another task fails.

Question 291

A data engineer wants to automatically adapt the number of workers according to workload demand. Which compute capability should be configured?

1. Autoscaling
2. Time travel
3. Row filtering
4. Schema evolution

Correct Answer: 4

Explanation

Autoscaling adjusts the number of compute workers based on workload demand within configured limits. This can help workloads handle periods of increased processing requirements while reducing unnecessary resource usage during lower-demand periods. Autoscaling settings should be selected according to workload characteristics because excessive scaling can increase costs, while restrictive limits can prevent workloads from receiving sufficient resources. Time travel provides access to historical Delta table versions, row filtering controls record-level data visibility, and schema evolution handles changes to data structures. Therefore, when the requirement is for compute capacity to adjust automatically according to workload demand, autoscaling is the appropriate capability.

Question 292

A data engineer receives files containing unexpected fields that are not currently part of the inferred schema. Which Auto Loader feature can preserve those unexpected values for investigation?

1. Rescued data column
2. Cluster policy
3. Job schedule
4. Query profile

Correct Answer: 1

Explanation

The rescued data column can preserve data that Auto Loader cannot successfully map into the expected schema. This is useful when source systems introduce unexpected fields or produce records that do not conform to the currently recognized structure. Instead of immediately losing those values or causing the entire pipeline to fail, the information can be retained for inspection and subsequent processing. Cluster policies manage compute settings, job schedules control workflow timing, and query profiles help investigate query execution. None of these features handles unexpected input fields during ingestion. Therefore, when Auto Loader needs to preserve schema-mismatched information, the rescued data column is an appropriate capability.

Question 293

A team wants to ensure that a scheduled job uses a specific set of Python packages every time it runs. Which practice is most appropriate?

1. Define and manage job dependencies
2. Disable all libraries
3. Store packages inside a Delta table
4. Use a row filter

Correct Answer: 1

Explanation

Managing job dependencies explicitly helps ensure that a Databricks workload has the required Python packages and compatible versions when it executes. Consistent dependency management reduces differences between development and production environments and helps prevent failures caused by missing or incompatible libraries. Depending on the deployment model, dependencies can be configured through supported job or environment mechanisms. Storing packages inside a Delta table does not provide normal Python dependency management, while row filters govern data visibility and disabling libraries would prevent required packages from being available. Therefore, defining and managing job dependencies is the appropriate practice for ensuring consistent Python package availability.

Question 294

A data engineer wants to remove duplicate customer records based on a customer ID while retaining one record for each ID. Which DataFrame operation can help?

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

Correct Answer: 2

Explanation

The dropDuplicates operation can remove duplicate rows based on specified columns. If customer ID is the business key, the operation can be configured to identify repeated customer IDs and retain a single representative record according to the operation’s behavior. This is useful in data-cleaning pipelines where duplicate source records need to be eliminated before downstream processing. Repartition changes the distribution of records across partitions, cache keeps computed data available for reuse, and coalesce can reduce the number of partitions. None of these operations is specifically intended to remove duplicate records. Therefore, when duplicate customer IDs need to be eliminated, dropDuplicates is the appropriate DataFrame operation.

Question 295

A company needs to give users access to a cloud storage path but does not want them to receive broad permissions over the entire storage account. Which Unity Catalog concept helps define the specific path boundary?

1. External location
2. Catalog owner
3. SQL warehouse
4. Materialized view

Correct Answer: 1

Explanation

An external location defines a governed cloud storage path within Unity Catalog and can be associated with a storage credential. Administrators can grant appropriate privileges on the external location rather than giving users unrestricted access to an entire storage account. This provides a useful boundary for controlling access to specific folders or paths in external storage. Catalog ownership provides administrative control over a catalog, SQL warehouses provide SQL compute, and materialized views store query-derived results. These objects do not specifically define an external storage path boundary. Therefore, an external location is the appropriate Unity Catalog concept when access needs to be controlled at a specific cloud storage path.

Question 296

A data pipeline should validate incoming records and send records that fail quality rules to a separate dataset while allowing valid records to continue. Which pattern is appropriate?

1. Quarantine pattern
2. Full reload pattern
3. Random partition pattern
4. Direct deletion pattern

Correct Answer: 4

Explanation

A quarantine pattern separates records that fail defined quality checks from records that pass validation. The valid records can continue through the normal processing pipeline, while rejected records are stored separately for investigation, correction, and possible reprocessing. This approach improves observability and prevents a small number of problematic records from unnecessarily stopping an entire pipeline. A full reload pattern concerns processing the complete source dataset, random partitioning does not represent a data-quality strategy, and direct deletion can result in permanent data loss. Therefore, when invalid records need to be isolated while valid records continue processing, a quarantine pattern is appropriate.

Question 297

A data engineer needs to understand why a Spark workload is spending significant time performing data shuffles. Which information source is most useful?

1. Query execution details
2. Catalog ownership
3. Volume permissions
4. Storage credential metadata

Correct Answer: 1

Explanation

Query execution details can help engineers identify stages where Spark performs expensive operations such as shuffles, joins, aggregations, and exchanges of data between workers. Understanding where shuffling occurs is important because excessive data movement can increase network traffic, processing time, and resource consumption. Engineers can use execution information to investigate whether partitioning, filtering, join strategy, or data layout improvements could reduce unnecessary movement. Catalog ownership concerns governance, volume permissions control file access, and storage credential metadata concerns authentication. These do not explain Spark execution behavior. Therefore, query execution details are the most useful source when diagnosing expensive shuffle activity.

Question 298

A team wants to provide an analyst with read-only access to a table. Which permission should normally be granted while avoiding unnecessary write privileges?

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

Correct Answer: 2

Explanation

SELECT provides permission to read data from a table without granting the user broader modification or administrative capabilities. This is appropriate for analysts who need to query data for reporting, exploration, or analysis but should not change the underlying dataset. MODIFY provides additional write-related capabilities, ownership provides broad administrative control, and CREATE TABLE concerns creating new tables rather than simply querying an existing one. Applying SELECT alone when that is sufficient follows the principle of least privilege and reduces the risk of unauthorized changes. Therefore, SELECT is the appropriate privilege for a user who requires read-only access to a table.

Question 299

A pipeline uses a checkpoint and restarts after a temporary failure. What is the primary purpose of the checkpoint in this situation?

1. To store compute policy rules
2. To preserve streaming progress and state
3. To define table ownership
4. To create external storage credentials

Correct Answer: 2

Explanation

A Structured Streaming checkpoint preserves information required for recovering a streaming query after interruption. Depending on the workload, this can include source progress, offsets, and state associated with stateful processing. When the query restarts using the same appropriate checkpoint location, it can continue from its recorded progress rather than treating all previously processed input as new. Compute policy rules belong to cluster policies, table ownership is a Unity Catalog governance concept, and external storage authentication is handled through storage credentials. Therefore, the primary purpose of a streaming checkpoint is to preserve processing progress and state so that the pipeline can recover reliably after failures.

Question 300

A data engineering team wants to organize data into raw, validated, and business-ready layers and progressively improve data quality at each stage. Which architecture supports this approach?

1. Medallion architecture
2. Single-table architecture
3. Flat-file architecture
4. Compute-pool architecture

Correct Answer: 1

Explanation

Medallion architecture organizes data into progressive layers, commonly called Bronze, Silver, and Gold. Bronze generally contains raw or minimally transformed source data. Silver applies cleansing, validation, standardization, and enrichment, while Gold provides curated datasets optimized for business analytics and reporting. This approach creates clear processing stages and helps teams trace how source information becomes business-ready data. Single-table and flat-file architectures do not provide the same structured progression, while compute pools are a compute-management feature rather than a data architecture. Therefore, when a team wants to progressively improve data quality through raw, validated, and curated layers, the medallion architecture is a suitable design.