DP-900: Microsoft Azure Data Fundamentals
This exam measures knowledge of core data concepts and related Microsoft Azure data services. It is intended for candidates who are beginning to work with data in the cloud and who are familiar with relational and non-relational data concepts. Candidates should also understand different types of data workloads, such as transactional and analytical workloads. The exam can serve as preparation for other Azure data-focused certifications.
What you get with membership
- The full DP-900 question bank with detailed explanations
- Readiness tracking by objective so you know when you're ready
- Access to every other exam in the Certavo library
- Content kept in step with the latest exam objectives
DP-900 exam objectives
The objective domains the DP-900 exam is built from.
Describe core data concepts
27.5% of the examDescribe an analytics workload on Azure
27.5% of the examIdentify considerations for relational data on Azure
22.5% of the examDescribe considerations for working with non-relational data on Azure
17.5% of the examDP-900 practice questions and answers
11 free sample questions from the DP-900 bank, with the correct answer and a full explanation for each. These are original questions written to the Microsoft objective domains — not real exam content.
- Question 1Describe an analytics workload on Azure
Jamal, a data engineer, must run existing Apache Spark notebooks that merge Delta Lake tables, execute PySpark feature-engineering code, and process both scheduled batch jobs and Structured Streaming inputs. The platform should provide collaborative notebooks and managed Spark clusters rather than a SQL-only warehouse. Which Azure service best fits the workload?
- AAzure Databricks workspace
- BAzure Stream Analytics job
- CAzure Synapse Analytics dedicated SQL pool
- DAzure Machine Learning managed online endpoint
Show answer and explanation
Correct answer
- Azure Databricks workspace
Explanation
Azure Databricks is a managed Apache Spark-based analytics platform for data engineering, data science, and machine learning. It supports collaborative notebooks, managed clusters, Delta Lake workloads, batch jobs, and Structured Streaming.
- Azure Stream Analytics is optimized for SQL-like stream processing; it is not a collaborative Spark notebook platform for PySpark and Delta Lake engineering.
- Azure Synapse Analytics dedicated SQL pool is a provisioned MPP SQL warehouse, not the Spark notebook and Structured Streaming environment requested.
- Azure Machine Learning supports model training and MLOps, but it is not primarily the managed Spark data-engineering workspace described in the scenario.
References:
- Question 2Describe an analytics workload on Azure
Several proposed workloads are being categorized before an Azure architecture review. Which two activities are examples of analytics workloads?
- AAggregating clickstream records to identify purchasing patterns
- BRunning sentiment analysis on customer comments
- CHosting a stateless web API with continuous deployment
- DRunning a parallel fluid dynamics simulation on compute nodes
Show answer and explanation
Correct answers
- Aggregating clickstream records to identify purchasing patterns
- Running sentiment analysis on customer comments
Explanation
Analytics workloads involve systematic analysis of data to discover, interpret, and communicate insights. Aggregating clickstream behavior and analyzing customer sentiment are both examples of deriving insight from data.
- Hosting a stateless web API is a cloud-native application workload rather than an analytics workload.
- Running a parallel fluid dynamics simulation is an example of a high-performance computing workload, not a typical business analytics workload.
References:
- Question 3Describe an analytics workload on Azure
Each night, supplier inventory files arrive as CSV exports. A scheduled job cleans the files and produces a report for managers the next morning. The workload does not require continuous event processing. Which processing pattern does this describe?
- ABatch analytics
- BReal-time stream analytics
- COnline transaction processing
- DHigh-performance computing
Show answer and explanation
Correct answer
- Batch analytics
Explanation
This is batch analytics. Data arrives as files on a schedule, is processed as a discrete set, and the report is produced later. Continuous stream processing is not required.
- Real-time stream analytics is used when events arrive continuously and results must be generated with low latency.
- Online transaction processing is used for day-to-day operational transactions, such as inserting and updating orders.
- High-performance computing focuses on parallel compute-intensive simulations or analysis, not scheduled business reporting from files.
References:
- Question 4Describe core data concepts
Quarter-end risk reports must scan 1.2 billion append-only policy snapshot rows and compute aggregates by date, region, and product line. Individual snapshots are not updated after they are written. Which Azure data platform design best fits this workload?
- AAn Azure Synapse dedicated SQL pool with clustered columnstore tables for large scans
- BAn Azure SQL Database normalized OLTP schema for many single-row updates
- CAn Azure Cache for Redis database with one key for each policy snapshot
- DAn Azure Blob Storage container with JSON files and no query engine
Show answer and explanation
Correct answer
- An Azure Synapse dedicated SQL pool with clustered columnstore tables for large scans
Explanation
An Azure Synapse Analytics dedicated SQL pool is designed for large-scale analytical processing. Clustered columnstore tables are well suited to large scans and aggregations because they store data by column and provide compression for analytical workloads.
- Azure SQL Database can support relational OLTP workloads, but a normalized schema optimized for many short row-level updates is not the best fit for billion-row analytical scans.
- Azure Cache for Redis is an in-memory cache, not a durable analytical platform for quarter-end reporting.
- Azure Blob Storage can store files, but storage alone does not provide the MPP query engine needed for the stated aggregations.
References:
- Question 5Describe core data concepts
You are selecting a store for feature flags in a multi-tenant web app. The app reads a few properties by tenant ID and flag name, does not join to other tables, and must use a simple low-cost Azure service rather than a globally distributed database. Which model should you choose?
- AAzure Table Storage table with tenant ID as `PartitionKey` and flag name as `RowKey`
- BAzure SQL Database tables normalized with parent and child relationships
- CAzure Cosmos DB for NoSQL with multi-region writes and custom indexing
- DAzure Blob Storage with one append blob for each tenant's feature flags
Show answer and explanation
Correct answer
- Azure Table Storage table with tenant ID as `PartitionKey` and flag name as `RowKey`
Explanation
Azure Table Storage is a NoSQL key/attribute store. Using
PartitionKeyfor the tenant andRowKeyfor the flag name supports efficient point lookups for simple entities without requiring relational joins.- Azure SQL Database is a relational engine and would work technically, but the scenario asks for a simple key/attribute model without joins.
- Azure Cosmos DB for NoSQL supports JSON documents and global distribution, but the requirement specifically avoids a globally distributed database for this simple, low-cost lookup pattern.
- Azure Blob Storage is object storage. It does not provide table-style
PartitionKeyandRowKeylookups over individual feature-flag entities.
References:
- Question 6Identify considerations for relational data on Azure
Woodgrove Bank stores regulatory trade records in a single Azure SQL Database. Retention rules are expected to increase the database size to about 18 TB. The architecture must remain a single managed database, avoid application-level sharding, and keep backup and restore operations efficient as the database grows. Which Azure SQL Database service tier should be selected?
- AHyperscale
- BBusiness Critical
- CGeneral Purpose
- DPremium
Show answer and explanation
Correct answer
- Hyperscale
Explanation
The Hyperscale service tier is designed for very large Azure SQL Database workloads, supports databases up to 100 TB, and uses a storage architecture that supports fast backup and restore operations.
- Business Critical is optimized for low-latency OLTP with local SSD storage and replicas, but it is not the tier intended for an 18-TB single database.
- General Purpose is a balanced tier for many workloads, but it does not provide Hyperscale capacity for very large single databases.
- Premium is a DTU-based tier and does not satisfy the large single-database growth requirement as directly as Hyperscale.
References:
- Question 7Identify considerations for relational data on Azure
Adventure Works is classifying platforms for two new operational applications. Both applications require a fully managed database engine, no virtual machine operating system administration, relational tables, joins, constraints, and transactions. One application uses T-SQL, and the other uses MySQL client libraries. Which TWO Azure services meet these requirements?
- AAzure SQL Database
- BAzure Database for MySQL flexible server
- CSQL Server on Azure Virtual Machines
- DAzure Cosmos DB for NoSQL
- EAzure Synapse Analytics dedicated SQL pool
Show answer and explanation
Correct answers
- Azure SQL Database
- Azure Database for MySQL flexible server
Explanation
Azure SQL Database and Azure Database for MySQL flexible server are managed relational database services that fit the stated operational application requirements and match the required SQL dialect or client protocol.
- SQL Server on Azure Virtual Machines can host relational SQL Server workloads, but it requires virtual machine and operating system administration.
- Azure Cosmos DB for NoSQL is a NoSQL database service; its SQL query syntax is not the same as a relational engine with foreign-key constraints and relational joins.
- Azure Synapse Analytics dedicated SQL pool is a relational analytics service based on a massively parallel processing architecture, not the primary choice for OLTP application transactions.
References:
- Question 8Identify considerations for relational data on Azure
Audit reports show inconsistent customer contact data after support agents edit phone numbers in several screens. The current design stores repeated values as comma-separated strings, and updates do not reliably keep related order records consistent. Which modeling change best addresses the issue in a relational Azure database?
- ACreate separate Customer, PhoneNumber, and Order tables with primary keys, foreign keys, and transactional updates.
- BKeep the comma-separated phone numbers in one table and split the strings when reports run.
- CStore each order as a standalone JSON document and update duplicate customer data asynchronously.
- DMove customer attachments to Azure Blob Storage and use a search index for the metadata.
Show answer and explanation
Correct answer
- Create separate Customer, PhoneNumber, and Order tables with primary keys, foreign keys, and transactional updates.
Explanation
A normalized relational model with separate tables and primary-key to foreign-key relationships addresses repeating groups and helps enforce consistency. Transactions can be used to commit related changes together in Azure SQL Database.
- Keeping comma-separated values in one column preserves the repeating-group problem and makes referential integrity difficult to enforce.
- Storing each order as an independent JSON document can be useful for some NoSQL designs, but it does not use relational constraints to maintain shared customer contact data.
- Moving files to Blob Storage and indexing metadata addresses unstructured content search, not relational data consistency.
References:
- Question 9Identify considerations for relational data on Azure
Noor, a data modeler, is reviewing a ticketing schema before it is deployed to an Azure relational database. Tickets must reference existing customers, and the ticket row plus its audit row must be committed as one unit. Which TWO database features address these requirements?
- ADefine a foreign key from the Tickets table to the Customers table.
- BWrap the ticket insert and audit insert in a single transaction.
- CCreate a nonclustered index on the customer display name column.
- DCreate a read replica for the database and send audit inserts to the replica.
- EUse an Azure Data Factory pipeline to copy the ticket row after it is inserted.
Show answer and explanation
Correct answers
- Define a foreign key from the Tickets table to the Customers table.
- Wrap the ticket insert and audit insert in a single transaction.
Explanation
A foreign key constraint enforces that a child row, such as a ticket, references an existing parent row, such as a customer. A transaction allows related statements to commit or roll back as a single unit, which is required for the ticket row and audit row.
A nonclustered index can improve lookup performance, but it does not enforce parent-child validity. A read replica can help scale read workloads, but it does not make a multi-statement change atomic. Azure Data Factory can orchestrate data movement and transformation, but it is not the mechanism that enforces OLTP referential integrity inside a relational database.
References:
- Question 10Identify considerations for relational data on Azure
Which TWO characteristics indicate that a workload is a good fit for a relational database service on Azure?
- AThe data is organized into tables with rows and columns and queried by using SQL.
- BRelationships between entities must be enforced by using keys and constraints.
- CThe primary requirement is to stream unstructured binary files without a fixed schema.
- DEach stored item can use a different structure, and the database is not expected to enforce relationships.
- EThe application stores opaque telemetry payloads and never joins or updates related records.
Show answer and explanation
Correct answers
- The data is organized into tables with rows and columns and queried by using SQL.
- Relationships between entities must be enforced by using keys and constraints.
Explanation
Relational workloads commonly organize data into tables with rows and columns and use SQL to query the data. They also commonly need relationships and constraints, such as primary keys and foreign keys, to enforce data integrity.
Unstructured binary files are usually a better fit for object storage such as Azure Blob Storage. Data in which every item can have a different structure and relationships are not enforced by the database is often associated with nonrelational models. Opaque telemetry payloads with no joins or transactional requirements do not specifically require a relational database.
References:
- Question 11Identify considerations for relational data on Azure
Marek, an operations lead, is moving a SQL Server database that uses a third-party backup agent installed on Windows and requires sysadmin-level instance access during a phased migration. The team accepts managing Windows patches but wants Azure infrastructure rather than on-premises hardware. Which Azure option should he choose?
- ASQL Server on Azure Virtual Machines
- BAzure SQL Database single database
- CAzure SQL Managed Instance
- DAzure Database for PostgreSQL flexible server
Show answer and explanation
Correct answer
- SQL Server on Azure Virtual Machines
Explanation
SQL Server on Azure Virtual Machines is the correct choice when the workload requires operating system access, installed Windows agents, and SQL Server instance-level control. It provides Azure infrastructure while retaining IaaS management responsibilities.
Azure SQL Database is a managed database service and does not provide Windows access or SQL Server instance administration. Azure SQL Managed Instance provides broad SQL Server compatibility as a PaaS service, but it still does not provide operating system access for installing Windows agents or full sysadmin control. Azure Database for PostgreSQL flexible server is for PostgreSQL workloads, not SQL Server workloads.
References:
Membership includes 198 questions and explanations aligned to the DP-900 curriculum, including 9 case studies.
Other Microsoft certifications
Every one of these is included with the same membership as DP-900.
Microsoft 365 Copilot and Agent Administration Fundamentals
This exam is intended for candidates who are familiar with Microsoft 365, including core services, security, identity and access, data…
Microsoft Azure AI Fundamentals
This exam measures knowledge of machine learning and AI concepts and related Microsoft Azure services. It is intended for candidates…
Microsoft Azure Fundamentals
This exam is intended for technology professionals who want to demonstrate foundational knowledge of cloud concepts and Microsoft Azure.…
DP-900 exam FAQ
How many questions are on the DP-900 exam?+
The DP-900 (Microsoft Azure Data Fundamentals) exam has around 50 questions. Question counts vary slightly between exam forms, so treat this as the typical number rather than a guarantee.
How long is the DP-900 exam?+
You get 120 minutes for the DP-900 exam itself. Allow extra time at the test centre or for the online check-in process before the timer starts.
What level is DP-900?+
DP-900 is a Microsoft fundamentals-level certification, so it assumes no prior certification and is a common starting point.
Are there free DP-900 practice questions?+
Yes. 11 free DP-900 practice questions are on this page, each with the correct answer and a full explanation. The complete bank of 198 questions is included with membership.
Are these real DP-900 exam questions?+
No. Every question is original, written to match the published DP-900 objective domains and question styles. Real exam content is confidential, and reusing it would breach Microsoft's exam policies.