The first hurdle with the azure-dp-900 exam is figuring out exactly what it expects you to know. For a first cloud certification, the breadth of the syllabus can feel overwhelming — like walking into a new city without a map. DP-900 is an entry-level exam that sweeps across four broad areas of working with data, and each area asks you to think a bit differently. Sketch out the boundaries between those four areas first, and it becomes much easier to know where to start.
What a 45-Minute Exam Is Actually Checking
A routine physical checkup does not take all day. It runs through a handful of key indicators in a fixed, short window and gives you a result. DP-900 works the same way: its full name is Microsoft AZURE Data Fundamentals, the exam runs 45 minutes, includes 50 questions, and the passing score is 700 out of 1000.
True to its Fundamentals-level status, DP-900 has no prerequisite and no experience requirement, and it never expires once earned — no renewal to track. The official guide lives on the Microsoft Learn page (learn.microsoft.com/credentials/certifications/azure-data-fundamentals). Most questions do not test whether you memorized a service name; they test whether you can pick the right category of storage or analytics tool for a described situation.
Sorting Data Into Three Drawers
Moving apartments means sorting belongings into a wardrobe, a drawer, and a box, which makes everything easier to find later. DP-900 starts from a similar exercise: every piece of data is either structured, semi-structured, or unstructured.
Structured data lives in fixed rows and columns, like a spreadsheet. Semi-structured data, such as JSON, has some organization but does not fold neatly into a table. Unstructured data — images, video, audio — carries no predefined schema at all. On top of that split, the exam asks when file formats like CSV, JSON, Parquet, ORC, and Avro each make sense. For the deeper breakdown of this classification and the file-format tradeoffs, the Core Data Concepts post covers that ground in detail.
People Who Use the Same Data Differently
Picture a convenience store's sales data. The cashier cares whether the payment happening right now went through correctly. Someone at headquarters cares which products sold well over the past month. That difference in perspective is exactly what separates OLTP (Online Transaction Processing) from OLAP (Online Analytical Processing).
OLTP is optimized for recording each transaction quickly and accurately as it happens. OLAP is optimized for pulling accumulated data together to spot broader trends. Layered on top, the exam covers what a database administrator, a data engineer, and a data analyst are each responsible for: the administrator keeps systems running, the engineer builds pipelines that move and reshape data, and the analyst extracts meaning from what has been collected. To practice telling workloads and roles apart, the Data Workloads and Roles post is the place to go.
Data That Plays by the Rules of a Table
Think of a library checkout card, where the name, checkout date, and due date each land in a specific, predefined slot. Data like that belongs in a relational database.
Relational databases use normalization to cut duplication, define relationships between tables with primary and foreign keys, and rely on SQL to query records. DP-900 also checks whether you understand how indexes speed up lookups, and what views and stored procedures do. On Azure that world includes Azure SQL Database, Azure SQL Managed Instance, and Azure Database for MySQL, PostgreSQL, and MariaDB — questions often ask which fits a given scenario. The Relational Data Fundamentals post walks through normalization and service differences in more depth.
Data That Refuses to Fit in a Table
Closet drawers hold clothes just fine, but a bicycle or camping gear belongs in a storage room instead. Some data is genuinely better off stored flexibly than forced into a rigid schema.
Azure Blob Storage holds unstructured data like images, video, and log files. Azure File Storage migrates existing file shares to the cloud. Azure Table Storage stores loosely-schemed key-value data. Azure Cosmos DB goes further as a globally distributed database that automatically replicates data across regions, supporting five APIs — Core, MongoDB, Cassandra, Gremlin, and Table. Partition key choice has a major impact on performance, and throughput is measured in RUs (Request Units). For a closer look at how these services differ, the Non-Relational Data Fundamentals post is a solid starting point.
From a Pile of Data to a Finished Picture
Coffee beans go through harvesting, roasting, packaging, and distribution before reaching a store shelf. Data follows a similar journey: collected through ETL or ELT, stored, analyzed, then turned into something visual.
Azure Synapse Analytics and Microsoft Fabric are unified platforms for pulling together and analyzing large volumes of data. Azure Databricks is a Spark-based engine strong at large-scale processing and machine learning. Azure Data Factory builds pipelines that move and transform data from multiple sources. When data needs processing the moment it arrives — sensor readings or click logs — services like Azure Stream Analytics, Event Hubs, and IoT Hub come into play. At the final stage, the tool that turns results into something a person can understand is Power BI, where you build reports in Desktop, share through Service, and define calculations with DAX. Large-scale and real-time analytics get more depth in the Large-Scale Data Analytics Services and Real-Time Data Analytics posts, with visualization covered in the Power BI Essentials post.
Comparing All Four Areas at a Glance
Packing a suitcase with clothes, toiletries, and documents in separate pouches means never hunting around the airport for something. Keeping DP-900's four areas separated the same way helps you classify questions faster once you're in the seat.
| Area | Core Question | Representative Services | |:--|:--|:--| | Core Data Concepts | Is this data structured, semi-structured, or unstructured? | Concept-focused, no specific service | | Relational Data | Does this data belong in tables with defined relationships? | Azure SQL Database, Azure Database for MySQL | | Non-Relational Data | Does it need a flexible structure or global distribution? | Blob Storage, Azure Cosmos DB | | Analytics and Visualization | How do you move, analyze, and present accumulated data? | Synapse Analytics, Data Factory, Power BI |
Starting from scratch, review Core Data Concepts and Data Workloads and Roles first to build an eye for classifying data, then compare relational and non-relational storage, and finish with analytics and Power BI once the storage layer feels solid.
Exam Key Takeaways
"Fixed rows and columns, table-shaped" -- Structured Data "Some structure, like JSON, but doesn't fold into a table" -- Semi-structured Data "No predefined schema, like images or video" -- Unstructured Data "Records each transaction quickly and accurately as it happens" -- OLTP "Pulls accumulated data together to analyze patterns" -- OLAP "Normalization, primary/foreign keys, queried with SQL" -- Relational Database "Globally distributed database, replicated across regions, throughput measured in RUs" -- Azure Cosmos DB "Stores unstructured data like images, video, and logs" -- Azure Blob Storage "Pipeline that moves and transforms data from multiple sources" -- Azure Data Factory "Processes real-time streaming data as it arrives" -- Azure Stream Analytics "Build reports, share them, define calculations with DAX" -- Power BI "45 minutes, 50 questions, 700/1000, no renewal required" -- DP-900 exam format
OLTP = real-time transaction recording, OLAP = analysis of accumulated data, Azure Cosmos DB = globally distributed non-relational storage
With the big picture across all four areas in place, it's time to put that judgment to the test with the DP-900 practice exam.