CertifyCloudxAce every cloud certification.

DP-900 sample questions with answers

10 free practice questions for the Microsoft Azure Data Fundamentals exam. Try each one, then open the answer to see why the right option wins and every other option loses.

Question 1Describe core data concepts

A parcel network is choosing between Azure SQL Database and Azure Cosmos DB for one service. The service must accept writes from depots on three continents with single-digit millisecond latency, and each parcel record carries different attributes depending on the country of origin. Which datastore should be chosen, and why?

  1. A.

    Azure Cosmos DB, because relational tables cannot store more than a fixed number of columns per row

  2. B.

    Azure Cosmos DB, because it distributes writes across regions and stores documents whose properties differ per record

  3. C.

    Azure SQL Database, because a single logical server can be reached from any region over the public endpoint

  4. D.

    Azure SQL Database, because parcel records with differing attributes can be held in one text column and parsed by the application

Show answer

Answer: B

Multi-region writes at single-digit millisecond latency over records with differing attributes is the Cosmos DB case.

  • A. The obstacle is schema variation and write distribution, not any column-count limit.
  • B. Globally distributed writes with low latency over per-record varying documents is Cosmos DB's core use case.
  • C. Reachability is not latency; writers on other continents still pay the round trip to a single region.
  • D. Hiding varying attributes in a text column removes field-level querying and pushes parsing into the application.
Question 2Describe core data concepts

A two-person start-up keeps roughly 40 GB of flat customer-preference entities, each retrieved by a customer identifier. Traffic is modest and confined to one region, no secondary indexes are needed, and the founders want the lowest running cost for a store that still scales if the product succeeds. Which Azure datastore should they choose?

  1. A.

    Azure SQL Database in a tier sized for the projected peak of the next three years

  2. B.

    Azure Table storage

  3. C.

    Azure Files, with one small file written per customer preference entity

  4. D.

    Azure Cosmos DB for NoSQL, configured with provisioned throughput across two regions

Show answer

Answer: B

Flat entities fetched by key in one region with no secondary indexes is the cheapest fit for Azure Table storage.

  • A. Sizing relational compute for a future peak pays for capacity now and adds relational features the workload never uses.
  • B. Flat entities by key, single region, no secondary indexes is precisely Table storage's low-cost niche.
  • C. One tiny file per entity gives no query capability and performs poorly as the number of entities grows.
  • D. Provisioned throughput across two regions buys distribution and indexing the workload explicitly does not need.
Question 3Describe core data concepts

A pharmacy chain stores dispensing records as structured data in a single wide table. Product teams now want to record a handful of extra attributes that apply only to veterinary medicines, roughly forty rows in every million. The data team must keep the existing representation. What is the most likely consequence of adding those attributes as new columns?

  1. A.

    Queries that reference the existing columns will stop returning results until the whole table has been reloaded with the new column list applied to every historical row

  2. B.

    Almost every row will hold a null in the new columns, because the fixed schema applies to all rows whether or not the attribute is relevant

  3. C.

    The table will automatically convert to a semi-structured representation once optional attributes exist

  4. D.

    The load will begin rejecting all human-medicine records because they omit the new columns

Show answer

Answer: B

A fixed schema applies to every row, so attributes relevant to a tiny minority of rows leave nulls in all the others.

  • A. Existing columns and the queries over them are unaffected by an added column, and no reload is required.
  • B. The schema is declared per table, so a rarely relevant attribute produces nulls on nearly every row.
  • C. Optional columns are ordinary in structured data; nothing about nulls converts a table into a semi-structured store.
  • D. Adding a nullable column does not invalidate existing records; they simply store null in it.
Question 4Describe core data concepts

An estate agency is designing a lettings system. Tenancy agreements, rent schedules and tenant records must be joined and updated transactionally, and each tenancy also has a folder of photographs and signed documents. Which arrangement should the agency adopt?

  1. A.

    Hold everything in Azure SQL Database, storing the photographs and signed documents in binary columns of the tenancy table

  2. B.

    Hold the tenancy records in Azure Table storage and the photographs in Azure Files, mounting the share from the web application

  3. C.

    Hold the tenancy records in Azure SQL Database and the photographs and documents in Azure Blob Storage, with each record storing the address of its files

  4. D.

    Hold everything in Azure Blob Storage, writing each tenancy record as a JSON file next to its photographs and documents

Show answer

Answer: C

Use a relational database for the records that need joins and transactions and object storage for the files, linking one to the other.

  • A. Binary columns inflate the database and pay relational prices for content the engine cannot query.
  • B. Table storage enforces no relationships between entities, and a mounted share is unnecessary for files a browser can fetch.
  • C. It matches each workload to the store built for it and links them with a stored address.
  • D. Files in a container provide no joins, no enforced relationships and no transactional updates.
Question 5Describe core data concepts

A travel aggregator exchanges several million availability messages a day with its partners. The current messages are XML, and the team wants to keep a human-readable, self-describing, hierarchical payload while reducing the number of bytes sent over the network for the same content. Which change should the team make?

  1. A.

    Keep XML and store the messages in a relational table, one message per row in a character column

  2. B.

    Move the messages to JSON, which represents the same hierarchy in a less verbose notation

  3. C.

    Move the messages to fixed-width text, allocating a set number of characters to each element in the current schema

  4. D.

    Move the messages to Parquet, and have each partner extract the columns they need from every message file

Show answer

Answer: B

JSON expresses the same hierarchy as XML with far less markup, so it stays human-readable and self-describing while sending fewer bytes.

  • A. Storing the same XML in a table changes nothing about the bytes exchanged with partners.
  • B. JSON carries the same hierarchy and self-description in a notably less verbose notation, cutting payload size.
  • C. Fixed-width text cannot express nesting or optional elements and is not self-describing.
  • D. Parquet is binary and columnar, suited to analytical files rather than to individual readable messages in flight.
Question 6Describe core data concepts

A regional airline is sorting five workloads. Which two should be placed in a managed relational database rather than in a non-relational or file store? Each correct answer presents a complete solution. (Choose TWO.)

Choose 2.

  1. A.

    Ticket sales, where a seat reservation and a payment must be committed together

  2. B.

    Engine sensor files landed hourly for later analysis by Spark

  3. C.

    A cache of recently rendered timetable pages fetched by one key

  4. D.

    Crew rostering, where a shift must reference an existing employee and hours are totalled across tables

  5. E.

    Cabin safety videos played on seat-back screens

Show answer

Answer: A, D

Rostering needs enforced references and cross-table aggregation, and ticket sales need a joint commit, which are both relational requirements.

  • A. Committing a reservation and a payment together requires a multi-row transaction.
  • B. Files landed for later bulk analysis belong in a data lake rather than an operational database.
  • C. Opaque values fetched by one key is the key-value pattern, where relational features add cost and latency.
  • D. Enforced references between shifts and employees plus cross-table aggregation are relational strengths.
  • E. Video is unstructured content and belongs in object storage, not in a relational table.
Question 7Describe core data concepts

A charity wants to answer questions such as which volunteers are connected to a given donor through at most four intermediate people, and the depth of those chains is not known in advance. Which type of data store is designed for this?

  1. A.

    A graph store, which holds entities as nodes and the relationships between them as edges

  2. B.

    A key-value store, which returns any related record instantly once its key is known

  3. C.

    A column family store, which keeps the related people together inside one column family per donor

  4. D.

    A file store, which keeps one file per volunteer and lets a search index resolve the connections between them

Show answer

Answer: A

Questions about chains of relationships of unknown depth are what graph stores, with nodes and edges, are built to answer.

  • A. Nodes and edges make variable-depth relationship traversal the store's native operation.
  • B. Key lookup presumes you already know the key, which is the part the question cannot answer in advance.
  • C. Column families group columns inside a row and provide no mechanism for following links between entities.
  • D. A search index finds text matches; it does not traverse a chain of relationships of unknown length.
Question 8Describe core data concepts

A retailer currently keeps six years of till receipts in the lake as uncompressed CSV and is considering Parquet instead. Which two benefits would the change deliver? Each correct answer presents a complete solution. (Choose TWO.)

Choose 2.

  1. A.

    Anyone can open the files in a text editor and read them unaided

  2. B.

    The files gain a transaction log that lets a reader see the table as it stood yesterday

  3. C.

    Individual rows can be updated in place without rewriting any file

  4. D.

    Queries that read a few columns avoid reading the rest of each row

  5. E.

    The files occupy less space, because storing like values together compresses well

Show answer

Answer: D, E

Parquet's columnar layout lets a query read only the columns it needs and compresses far better than row-oriented text.

  • A. Parquet is binary, so human readability is the property being traded away by the move.
  • B. Versioning and time travel come from Delta Lake's transaction log, not from Parquet files on their own.
  • C. Parquet files are written once; updating rows means rewriting files unless a table format such as Delta Lake is used.
  • D. Columnar storage means a query reads only the column chunks it needs, skipping the remaining columns.
  • E. Storing like values together compresses far better than row-interleaved text, so the files occupy much less space.
Question 9Describe core data concepts

A garden design studio is closing its office server. The server holds a small desktop database of client projects that the studio wants to keep querying with SQL, and a 3 TB library of site photographs that designers currently open from a mapped network drive on both Windows and Mac laptops. Which pair of Azure datastores should the studio adopt?

  1. A.

    Azure Blob Storage for both, keeping the client projects as exported spreadsheet files beside the photographs

  2. B.

    Azure Cosmos DB for NoSQL for the client projects and Azure Data Lake Storage Gen2 for the photograph library

  3. C.

    Azure SQL Database for the client projects and Azure Files for the photograph library

  4. D.

    Azure Table storage for the client projects and Azure Blob Storage for the photograph library

Show answer

Answer: C

The project records keep their SQL querying in Azure SQL Database, and a library opened from a mapped drive needs the mountable share that Azure Files provides.

  • A. Exporting the projects to files abandons querying entirely, which the studio explicitly wants to keep.
  • B. Cosmos DB changes how the records are queried, and a data lake is for analytics rather than a designers' working share.
  • C. It preserves SQL querying for the records and provides a mountable share for the photograph library.
  • D. Table storage does not offer SQL querying of related project records, and Blob Storage cannot be mapped as a drive.
Question 10Describe core data concepts

A finance team exports ledger entries to comma-separated files and loads them into a reporting tool. Account codes such as 00480 arrive in the report as 480, and a column of dates is sometimes interpreted as day-month and sometimes as month-day. Which property of the file format is responsible?

  1. A.

    Comma-separated files store a checksum per row, and a mismatch causes the reader to fall back to a default type

  2. B.

    Delimited text supports only the numeric type, so text values are silently converted when the file is written

  3. C.

    Delimited text records characters only and carries no data types, so each reader decides how to interpret a value

  4. D.

    The file is columnar, so values in the same column are compressed together and lose their original precision

Show answer

Answer: C

CSV holds printable characters with no type information, so every reader must guess whether a value is text, a number or a date.

  • A. Delimited text has no checksum mechanism and no type fallback behaviour of this kind.
  • B. CSV stores every value as characters; it does not restrict content to a numeric type or convert on write.
  • C. Untyped characters force each reader to infer types, which is what strips leading zeros and flips date order.
  • D. CSV is row-oriented plain text, not columnar, and compression does not alter stored values in any format.

Keep going with 507 more DP-900 questions

Free papers every day, in the real exam formats, with progress by exam domain. Unlock every paper and timed mock exam when you are ready.

Practise freeSee all DP-900 papers
DP-900 sample questions with answers (10 free) · CertifyCloudx