Microsoft Azure & FabricLive · 102 questions · 11 notes

DP-900: Microsoft Azure Data Fundamentals

Free practice exam, study notes with architecture diagrams and the official blueprint for Microsoft Azure Data Fundamentals.

Passing score
700 / 1000
Domains
4
Exam code
DP-900

Official DP-900 exam blueprint

Domains and weights from Microsoft Azure & Fabric's exam guide.

DomainWeight
Describe core data concepts25–30%
Identify considerations for relational data on Azure20–25%
Describe considerations for working with non-relational data on Azure15–20%
Describe an analytics workload on Azure25–30%

Free DP-900 sample questions

The first questions of the free diagnostic. Pick your answer, then check it. Why every option is right or wrong, and the What-If twist, open in the workspace.

  1. Question 1 · Describe ways to represent data

    Last night's load of the order feed into a fixed-column SQL table failed with a column mismatch error after a release added giftWrap to some line items. New attributes arrive with almost every release, and the analytics team must keep loading orders without first agreeing a fixed column list. Every value still has to be queryable by its field name. An order looks like this.

    JSON
    {
      "id": "1001",
      "customer": "Contoso",
      "items": [
        { "sku": "A-100", "qty": 2 },
        { "sku": "B-200", "qty": 1, "giftWrap": true }
      ]
    }

    How should the team classify this order data before redesigning the load?

    1. AStructured data
    2. BUnstructured data
    3. CBinary large object (BLOB) data
    4. DSemi-structured data
    Check the answer

    Answer: D. Semi-structured data such as JSON carries its own field names, so records can vary and gain fields while every value stays addressable by key.

    Official docs: Microsoft Learn: Identify data formats
    Unlock Full 4-Option Autopsy & What-If Twist in Workspace
  2. Question 2 · Describe relational concepts

    A marketing team must email only customers who have placed at least one order, and the list must not include customers without orders. An analyst runs the query below against 50,000 customers and 210,000 orders before the campaign deadline.

    SQL
    SELECT c.CustomerName, o.OrderDate
    FROM dbo.Customers AS c
    INNER JOIN dbo.Orders AS o
      ON o.CustomerID = c.CustomerID;

    What does this query return?

    1. AEvery customer, with a NULL order date for customers who have no orders
    2. BOnly customers who have at least one matching order, once per order
    3. COnly orders that have no matching customer
    4. DEvery combination of every customer with every order
    Check the answer

    Answer: B. An INNER JOIN returns only matching rows; a LEFT OUTER JOIN keeps every row from the left table and returns NULLs where nothing matches.

    Official docs: Microsoft Learn: Joins (SQL Server)
    Unlock Full 4-Option Autopsy & What-If Twist in Workspace
  3. Question 3 · Describe capabilities and features of Azure Cosmos DB

    An orders container in Cosmos DB must spread 400 million items evenly across physical partitions, and every transactional batch may only touch the items of one customer. The developer is creating the container for items like this:

    JSON
    { "id": "ord-9921",
      "customerId": "c-1187",
      "total": 129.90,
      "status": "shipped" }

    Which key setting defines that logical boundary?

    1. AA partition key of /customerId
    2. BA primary key constraint
    3. CA foreign key to a customers container
    4. DAn indexing policy path
    Check the answer

    Answer: A. The partition key groups items into logical partitions for scale-out and scopes transactional batches; the indexing policy controls indexing cost.

    Official docs: Microsoft Learn: Partitioning and horizontal scaling in Azure Cosmos DB
    Unlock Full 4-Option Autopsy & What-If Twist in Workspace
Answer all 10 free DP-900 questions

DP-900 study notes: key concepts

What each note covers. The full notes, with summaries, examples and memory hooks, open with a free account.

Architecture flow · Describe ways to represent data

Choosing a representation and store for a new dataset
  • A new dataset arrives for a workload

      • Does every record share the same typed columns?

        • yes
          • Structured rows in Azure SQL Database

        • no
          • Do keys or tags label each value inside a record?

            • no
              • Unstructured files in Azure Blob Storage

            • yes
              • Are relationships between entities the main question?

                • yes
                  • Graph of vertices and edges in Cosmos DB for Apache Gremlin

                • no
                  • Semi-structured JSON items in Cosmos DB for NoSQL

Drawn fromMicrosoft Learn: Identify data formats

Describe ways to represent data

  • Structured data has a schema of typed columns defined before loading, and the database enforces it on write, so a row whose value has the wrong type is rejected.
  • Semi-structured data such as JSON or XML labels every value with a key or tag, so records can differ, nest arrays and gain fields without any schema change.
  • Unstructured data such as images, audio, video and free text has no internal fields and becomes queryable only after a service extracts metadata into a new dataset.
  • Graph data stores entities as vertices and relationships as edges with properties, so questions that follow several hops of relationships run as traversals instead of joins.
  • The representation decides the store: Azure SQL Database for structured rows, Cosmos DB for JSON, Blob Storage for files and the Gremlin API for graphs.

Identify options for data storage

  • Delimited text such as CSV is readable by every tool but stores no data types, so each reader must infer or declare the types of every column itself.
  • Avro is a row-based binary format that stores its schema in the file header, which suits streaming, serialization and writing one record at a time.
  • Parquet and ORC are columnar formats that compress each column together, so analytical queries read only the columns they need from very large files.
  • Delta Lake adds a transaction log to Parquet files, giving lakehouse tables ACID commits, schema enforcement and time travel to earlier versions.
  • Non-relational databases come as key-value, document, column-family and graph stores, and each is chosen by the way the application looks data up.

Describe common data workloads

  • OLTP handles many small, concurrent reads and writes on current data, committing each as an ACID transaction so balances and stock are never seen half-updated.
  • ACID means atomicity, consistency, isolation and durability, while BASE stores accept eventual consistency in exchange for availability and easier scale across regions.
  • OLAP aggregates large volumes of historical data in a warehouse or lakehouse that ETL or ELT pipelines load, so reports never slow down the transactional system.
  • A star schema joins one fact table of measures directly to denormalized dimension tables, so every slice needs a single join and scans efficiently.
  • A snowflake schema normalizes dimensions into further related tables, which removes repeated values but adds join depth to every query that slices by those attributes.

Identify roles and responsibilities for data workloads

  • A database administrator secures operational databases with least-privilege roles and firewall rules, keeps them available, and runs and tests backups and point-in-time restores.
  • A data engineer builds, schedules and monitors pipelines that ingest, clean and transform source data into lakes, lakehouses and warehouses for analysis.
  • A data analyst explores curated data, models it in Power BI, builds reports and dashboards, sets alerts and explains the resulting insights to decision makers.
  • Ownership follows the chain: ingestion problems go to the engineer, report and measure problems to the analyst, and security or recovery problems to the DBA.
  • Business users consume reports and dashboards, and software developers build applications, so neither role is the answer for pipeline or modelling work.

Describe relational concepts

  • A primary key uniquely identifies each row and cannot be NULL, and a foreign key in another table references it so the engine enforces referential integrity.
  • With the default NO ACTION rule, deleting a referenced row fails, while ON DELETE CASCADE deletes the referencing rows together with the parent row.
  • Normalization through 1NF, 2NF and 3NF stores each fact once, removing the insert, update and delete anomalies that repeated values cause in a table.
  • DDL statements define objects, DML statements read and change rows, and DCL statements such as GRANT, REVOKE and DENY manage who may run what.
  • A view is a stored query that holds no rows, an index speeds lookups without enforcing values, and a stored procedure packages reusable SQL logic.

Describe relational Azure data services

  • SQL Server on Azure Virtual Machines is IaaS, so you get operating system access for agents and SSRS but must patch, back up and protect it yourself.
  • Azure SQL Managed Instance keeps instance features such as SQL Server Agent, Service Broker and cross-database queries while Microsoft patches the operating system and engine.
  • Azure SQL Database is fully managed at database level, always on the latest stable engine, and is bought under either the DTU or the vCore purchasing model.
  • Serverless compute scales vCores with demand and auto-pauses when idle, billing only storage while paused, at the price of a resume delay on the next query.
  • Hyperscale separates compute from storage so one database grows to very large sizes, with readable replicas and fast snapshot-based restores.
Read the full DP-900 study notes

DP-900 FAQ

What is the passing score for DP-900?

The passing score for Microsoft Azure Data Fundamentals is 700 / 1000.

Which domains does DP-900 cover?

The official exam guide lists 4 domains: Describe core data concepts (25–30%); Identify considerations for relational data on Azure (20–25%); Describe considerations for working with non-relational data on Azure (15–20%); Describe an analytics workload on Azure (25–30%).

Is there a free DP-900 practice test?

Yes. Gyan AI's free diagnostic is 10 DP-900 questions with no sign-up, each with an explanation for every option and a link to the official documentation. A free account unlocks the full question set and study notes.

Are these the real DP-900 exam questions?

No. Real exam content is confidential under vendor exam agreements, and using leaked "dumps" can get a certification revoked. Gyan AI questions are original and written to the published skills outline.