Can You Answer These 5 DP-900 Questions on Describe core data concepts?

Practice Describe core data concepts for the Azure Data Fundamentals (DP-900) exam with the 5 questions below. It carries about 27.5% of the exam score. Every option is explained, so a wrong pick still teaches something.

Azure Data Fundamentals (DP-900) · Describe core data concepts · 5 questions with answers and explanations

🧭 What these questions look for

Questions about core data concepts are recognition tasks. A short description of a file, a data source, a processing job or a daily task asks you to name the format, data category, processing style or data role. Read the defining detail carefully: one feature, such as braces and name and value pairs, or a pipeline that fails overnight, usually settles it.

Watch for these while you work

• Sort data by how fixed its shape is: identical columns, flexible fields in JSON or XML, and media or documents are three different categories.
• Separate the data roles by daily work: access and backups belong to one role, building and watching pipelines to another.
• Batch processing works on complete datasets, so the trade-off to remember is the delay before results appear.

Question 1 of 5

A developer opens a data file in which objects are enclosed in braces, collections in square brackets, and each attribute is written as a name and a value separated by a colon. Which format is this?

AApache Avro, a row-based format that stores its records in binary blocks after a header that describes the data.

BComma-separated values (CSV), in which each row is a line of plain text and each field is separated by a comma.

CExtensible Markup Language (XML), which wraps every element and attribute in opening and closing angle-bracket tags.

DJavaScript Object Notation (JSON), a hierarchical format suited to both structured and semi-structured data.

Show answer

✅ Answer: D

Difficulty: applied

JavaScript Object Notation (JSON)
JSON uses a hierarchical document schema to define entities (objects) with multiple attributes.
• Objects are enclosed in braces { }.
• Collections (arrays) are enclosed in square brackets [ ].
• Attributes are name : value pairs separated by commas.
• An attribute's value can itself be an object or a collection, so data can nest.
That flexibility makes JSON good for both structured and semi-structured data.

Example
{"customers": [{"firstName": "Joe", "contact": [{"type": "email", "address": "joe@litware.com"}]}]}

How it compares
• XML expresses the same kind of hierarchy with tags in angle brackets and is more verbose.
• CSV is flat: one line per row, no nesting.

Why each option works or fails
• ✗ A. Apache Avro: Avro stores its records as binary blocks, which a developer cannot read as braces and name-value pairs; only its header is JSON.
• ✗ B. CSV: CSV has no objects or collections; it is flat text with one record per line and comma-separated fields.
• ✗ C. XML: XML uses angle-bracket tags such as <Customer>, not braces and colon-separated name-value pairs.
• ✓ D. JSON (correct): braces for objects, square brackets for collections and name-value pairs for attributes are the syntax of JSON, a hierarchical format for structured and semi-structured data.

Question 2 of 5

A new sales application needs read access to the customer database, and a departing contractor's sign-in must lose all permissions on it. Which data role is typically responsible for these tasks?

AA database administrator, who manages the security of the data in the database and grants or denies users access.

BAn AI engineer, who integrates large language models and machine learning features into applications and workflows.

CA data engineer, who builds and monitors the pipelines that move and transform data between systems for analysis.

DA data analyst, who explores the data to find trends and builds reports and visualizations that the sales team uses.

Show answer

✅ Answer: A

Difficulty: applied

The main data roles
Microsoft's DP-900 material describes these key roles (organizations may name or split them differently):
• Database administrator (DBA): designs, implements, maintains and operates on-premises and cloud database systems; responsible for availability, consistent performance and optimization, backup and recovery, and security: granting privileges over the data and granting or denying user access.
• Data engineer: designs and implements data workloads: ingestion pipelines, cleansing and transformation, and data stores for analytics; manages and monitors pipelines.
• Data analyst: explores and analyzes data, builds analytical models and creates reports and visualizations for decisions.
• AI engineer: the Learn module also lists this key role, which builds AI-powered features with large language models, machine learning pipelines and data sources; the exam study guide names only the three roles above.
• Other roles, such as data scientists and data architects, also work with data.

Example
A DBA creates a login for the new app with read-only permission on the customer tables, and removes the contractor's database permissions on his last day.

Why each option works or fails
• ✓ A. Database administrator (correct): managing database security, granting privileges and granting or denying user access, is a core DBA responsibility.
• ✗ B. AI engineer: AI engineers build AI-powered features with models and data sources; database permissions are outside that role.
• ✗ C. Data engineer: engineers build pipelines and analytical stores; permissions on an operational database are a DBA task.
• ✗ D. Data analyst: analysts consume data to produce insights and reports; they are not the people who grant or remove access to the database.

Question 3 of 5

A data team is sorting its sources into structured, semi-structured and unstructured data. Which TWO sources are unstructured data? (Select TWO)

AAn XML feed of supplier prices that uses nested elements and attributes to describe each product.

BA CSV export of the order table, with a header row and the same six columns on every line.

CJSON messages from a web API in which some optional fields appear only for some of the records.

DAudio recordings of support calls that are saved as MP3 files at the end of every conversation.

EPhotos of damaged goods that customers upload as JPEG files together with their return requests.

Show answer

✅ Answer: D, E

Difficulty: applied

Sorting data by structure
• Structured: a fixed schema, the same fields in every record, usually rows and columns. A table, or a CSV file whose lines all have the same columns.
• Semi-structured: some structure with variation allowed between records. JSON and XML documents, where fields and nesting can differ from one record to the next.
• Unstructured: no specific structure. Images, audio, video, documents and other binary files.

A quick test
Ask: "Can I name the fields inside one record?"
• Yes, and every record has the same fields: structured.
• Yes, but the fields can differ or nest: semi-structured.
• No, the content must be interpreted (seen, heard, read) first: unstructured.

Why each option works or fails
• ✗ A. XML price feed: tags and nested elements give the data a structure that can vary per product, so it is semi-structured.
• ✗ B. CSV export of a table: every line has the same six columns, which makes it structured data even though it is stored as a text file.
• ✗ C. JSON with optional fields: named fields that vary between records are the definition of semi-structured data.
• ✓ D. Call recordings (MP3) (correct): audio files have no internal field-and-value layout, so they are unstructured data.
• ✓ E. Photos (JPEG) (correct): images are raw binary content with no fields inside them, a standard example of unstructured data.

Question 4 of 5

A retailer runs a nightly job that reads all of the day's sales files, aggregates them and writes summary tables that reports use the next morning. What is the main drawback of this batch approach?

AThe job cannot aggregate data, because batch processing can only copy files from one location to another unchanged.

BThe job must keep every source file locked for writing all day, so the stores cannot record any sales until it finishes.

CThe job loses accuracy, because batch processing samples only a small part of each file to finish within its time window.

DResults can be several hours old, because the job processes the data only after a whole period has been collected.

Show answer

✅ Answer: D

Difficulty: applied

Batch processing
In batch processing, data is collected over a period and then processed together by a scheduled, often long-running job.
• Big data solutions use batch jobs to filter, aggregate and prepare large datasets for analysis: read source files, process them, write the output to new files or tables.
• Batch processing works on data at rest, usually stored in a data lake.

The trade-off
• Strengths: handles very large volumes efficiently and processes the complete dataset, so results are complete and accurate.
• Drawback: latency. If processing runs once a night or takes hours, query results are hours old.

Other workload types
Big data solutions can also include interactive exploration of data (ad hoc queries and notebooks), real-time processing of data in motion, and machine learning. Choosing among them depends on how fresh the answers must be.

Example
Sales made at 09:00 Monday appear in the summary tables at 06:00 Tuesday; for daily reporting that is fine, for live stock alerts it is not.

Why each option works or fails
• ✗ A. Cannot aggregate: filtering, aggregating and preparing data is exactly what batch jobs in big data solutions do.
• ✗ B. Locks source files all day: the job reads files that were already written; the point-of-sale systems keep recording sales in their own databases.
• ✗ C. Loses accuracy by sampling: batch processing reads the complete dataset and is usually the more accurate path; reduced accuracy is a trade-off of fast low-latency paths.
• ✓ D. Latency of hours (correct): a batch job runs after the period's data has been collected, so the summaries always lag behind the latest sales, which is the classic batch trade-off.

Question 5 of 5

Overnight data loads into the warehouse have started finishing late, and some mornings a load fails without anyone noticing until reports show old numbers. Whose responsibility is it to manage and monitor these loads?

AThe business users, because they decide which reports are important and should check that the figures look right each day.

BThe data analyst, because analysts are the first people to see the old numbers when they open their reports each morning.

CThe database administrator of the source system, because the source database is where the data that is being loaded is stored.

DThe data engineer, because data engineers own the management and monitoring of data pipelines so loads perform as expected.

Show answer

✅ Answer: D

Difficulty: applied

Owning the pipelines
Data engineers do not just build pipelines once. Microsoft's role description says they own the management and monitoring of data pipelines to ensure that data loads perform as expected.
That includes:
• Watching run times and failures, and alerting on them.
• Fixing transformation logic or source connection issues.
• Tuning loads that take too long.

Working with other roles
• The DBA of a source database may help if the source itself is slow or unavailable.
• The data analyst depends on fresh data and often reports the symptom.
• Neither of them owns the pipeline.

Example
The data engineer adds an alert when the nightly load fails or runs past 05:00, and moves a slow transformation step so the warehouse is ready before the 07:00 reports.

Why each option works or fails
• ✗ A. Business users: business users consume reports; they are not expected to monitor data loads.
• ✗ B. Data analyst: analysts may notice stale figures first, but noticing a symptom does not make them responsible for running and monitoring pipelines.
• ✗ C. Source database administrator: the DBA looks after the source database's availability and performance, not the integration pipelines that copy its data.
• ✓ D. Data engineer (correct): managing and monitoring data pipelines so that loads complete as expected is part of the data engineer's role.

🎯 Quick recap

• Braces, square brackets and name : value pairs = JSON, a hierarchical text format.
• Granting and revoking database access is a database administrator responsibility.
• Images, audio, video and documents are unstructured; JSON and XML are semi-structured; same-columns tables and CSV are structured.
• Batch processing handles large complete datasets accurately but its results lag (latency).
• Monitoring and managing data pipeline runs is a data engineer responsibility.

📊 This set at a glance

Exam: DP-900 · 45 questions · 45 minutes · passing score 700
This set: 5 applied · 2 scenario, 1 comparison, 1 concept check, 1 trade-off · 1 with more than one answer

🔗 Related practice sets

• All DP-900 practice sets
• Every exam on this blog

📚 Keep studying DP-900

More DP-900 questions: https://hyunhp.github.io/cert-bank/#exam=DP-900
Study guide: https://hyunhp.github.io/cert-bank/study.html#exam=DP-900

These are original practice questions, not real exam questions, and cert-bank is not affiliated with any certification vendor.

Original practice questions written for this blog, not exam dumps.