DP-900 Azure Data Fundamentals Study Guide

The DP-900 Exam

DP-900 Azure Data Fundamentals tests core data concepts and how Microsoft Azure services are used to build data solutions. It is the entry point to Azure data certifications and needs no advanced technical skills or hands-on implementation experience.

The exam has 40-60 questions and takes about 45 minutes. The passing score is 700 out of 1000. Deeper Azure certifications test implementation; DP-900 tests concepts.

Who Should Take DP-900?

DP-900 suits people at several career stages:

• Data analysts seeking to understand Azure's data platforms
• Business intelligence (BI) developers new to Azure
• Database administrators new to cloud-based data services
• Non-technical managers who need to understand data architectures
• Anyone planning to pursue advanced Azure data certifications like AI-300 (Machine Learning Operations Engineer, replaces the retired DP-100) or DP-700 (Fabric Data Engineer Associate, replaces the retired DP-203)

It also suits career changers moving into data, and people moving from on-premises systems to Azure.

Exam Difficulty and Comparison to AZ-900

DP-900 is at the same difficulty level as AZ-900 Azure Fundamentals. Both are conceptual exams for entry-level candidates without deep technical backgrounds. The exam does not ask you to write code, run complex queries or configure anything.

Instead, it asks which Azure service fits a given data problem, when to use a relational or non-relational database, and which analytics solution suits a use case. Passing depends on understanding and recognising services rather than hands-on skill.

DP-900 Exam Domains and Weighting

Microsoft splits DP-900 into four domains. The weights show where to spend study time.

Domains carry different weights, so dedicate time to each one in proportion to its share of the exam, and prioritise the largest domains first.

Domain 1: Describe Core Data Concepts (25-30%)

Understanding Data Fundamentals

The rest of the exam builds on this domain. Start by telling structured, semi-structured and unstructured data apart.

Structured data fits into tables with a defined schema, as in a relational database. Every row follows the same column definitions, which makes it efficient to query and analyse.

Semi-structured data has some organisation but no rigid schema. JSON and XML are the usual examples: their tags or keys describe the data without enforcing a strict structure.

Unstructured data has no predefined format. Images, video, audio files and text documents are unstructured, and they need their own processing and storage approaches.

Batch and Streaming Data Processing

Batch processing handles large volumes of data at scheduled intervals, typically overnight or during quiet periods. It suits historical analysis and loading a data warehouse.

Stream processing analyses data in real time as it arrives. Fraud detection and sensor monitoring need results immediately, so they use streaming. Streaming gives up some computational efficiency in exchange for real-time results.

Data Analysis and Visualisation

Data analysis means examining data to find patterns and answers, from simple aggregations to predictive modelling. Different audiences need different depths of analysis.

Visualisation presents the results as charts, graphs and dashboards, so executives and operational teams can read them quickly.

Domain 2: Identify Considerations for Relational Data on Azure (20-25%)

Relational Database Fundamentals

Relational databases store data in tables linked by relationships. Each table has columns (attributes) and rows (records). A primary key identifies each row, and foreign keys link tables together.

Normalisation reduces duplicated data by splitting it into logically organised tables. A well-normalised database uses less storage and keeps data consistent through referential constraints.

Relational databases are queried with Structured Query Language (SQL). For DP-900, know SELECT, INSERT, UPDATE and DELETE.

Azure SQL Database

Azure SQL Database is Microsoft's platform-as-a-service (PaaS) relational database. Patching, backups and infrastructure are handled for you, so teams spend their time on data instead of server administration.

It includes automatic scaling, built-in intelligence and security features. There are two deployment models: a single database suits an application with a predictable workload, and an elastic pool shares resources across several databases whose demand varies.

Backup and recovery are automatic. Microsoft keeps geo-redundant copies to protect against a regional outage, and point-in-time restore lets you recover to a specific moment if data is corrupted.

Azure SQL Managed Instance

Azure SQL Managed Instance sits between Azure SQL Database and SQL Server on a virtual machine. It is more compatible with on-premises SQL Server than SQL Database is, which makes migration easier for organisations with complex requirements.

Managed Instance supports instance-level features that SQL Database lacks, such as SQL Agent jobs and linked servers. It costs more than SQL Database and less than a virtual machine deployment.

Domain 3: Describe Considerations for Non-Relational Data on Azure (15-20%)

Azure Cosmos DB Overview

Azure Cosmos DB is a globally distributed, multi-model NoSQL database service. One service supports several APIs and data models.

Cosmos DB guarantees single-digit millisecond latency at scale and fails over automatically across Azure regions. It suits very large document collections and workloads that need to scale quickly.

Cosmos DB Data Models

The document model stores data as JSON documents with flexible schemas, queried through the SQL or MongoDB APIs. It suits web applications and content management systems.

The key-value model maps each unique key to a value. It is very simple, retrieves data quickly and works well for caching and session storage.

The graph model stores data as nodes and edges, which makes relationship queries straightforward. Social networks and recommendation engines use it.

The column-family model organises data by columns rather than rows, which speeds up analytical queries over large datasets. It suits time-series data and sensor applications.

Azure Blob Storage

Azure Blob Storage holds large amounts of unstructured data at low cost. Data sits in containers, and each container holds blobs of any file type.

Three access tiers control cost. The hot tier is for frequently used data with instant access. The cool tier costs less and suits data accessed monthly or less. The archive tier is cheapest, for data accessed rarely or never, and retrieval takes longer.

Domain 4: Describe an Analytics Workload on Azure (25-30%)

Data Warehousing with Azure Synapse Analytics

Azure Synapse Analytics combines big data and data warehouse features. It stores historical data for trend analysis and runs complex analytical queries across terabytes of data.

Synapse uses a massively parallel processing (MPP) architecture: each query is split across many nodes at once, so very large datasets are analysed quickly. Organisations use it for executive dashboards, financial analysis and business intelligence.

Big Data Processing with Azure Databricks

Azure Databricks is a managed Apache Spark environment for processing large datasets. Data scientists and engineers use it for machine learning, data preparation and complex transformations.

Databricks notebooks support Python, Scala, SQL and R. Clusters scale automatically as demand changes, and built-in connections to Azure Storage and SQL services make it easy to move data in and out.

ETL and ELT Pipelines

Extract-Transform-Load (ETL) extracts data from source systems, transforms it into a usable format and loads it into the target system. Because the transformation happens before loading, only clean data reaches downstream systems.

Extract-Load-Transform (ELT) loads raw data first and transforms it inside the target system. It suits cloud environments with plenty of compute, because cloud data warehouses run transformations more efficiently than traditional ETL tools.

Azure Data Factory

Azure Data Factory orchestrates data movement and transformation across hybrid environments. It connects to hundreds of on-premises and cloud data sources, and its visual pipeline designer lets non-developers build complex workflows.

Data Factory monitors pipeline runs, retries failed activities and alerts teams to problems. It works with Databricks, Synapse and other Azure services to build complete solutions.

Power BI for Visualisation

Power BI turns data into interactive visuals and dashboards that business users can explore without writing code. It connects to many data sources, including Azure SQL Database, Synapse and CSV files.

Dashboards show headline metrics for executives, and reports let analysts explore the detail. Power BI Premium allows sharing across an organisation without complicated licensing.

Study Plan and Timeline

Plan on 2-4 weeks of study for DP-900, depending on how much data knowledge you already have.

Week 1 is core data concepts: structured, semi-structured and unstructured data, the difference between batch and stream processing, and an overview of Azure's data services.

Week 2 is relational data. Study relational database principles in detail, learn how Azure SQL Database and Managed Instance differ, and practise deciding when to use each.

Week 3 is non-relational services and analytics: Cosmos DB data models, Blob Storage access tiers, and what Synapse, Databricks and Data Factory each do.

Week 4 is practice exams and review. Find your weak domains, study them again and sit full-length practice tests.

With little background in data concepts, add an extra week. With substantial data experience on other platforms, 2-3 weeks may be enough.

Key Study Concepts

These concepts come up most:

Service Categories: Know which services handle relational data, non-relational data and analytics. Sorting services into these groups avoids confusion in the exam.

Use Case Matching: Practise matching business scenarios to Azure services. Many exam questions describe a situation and ask for the best solution.

Scalability and Performance: Know which services scale automatically and which need manual management, and how latency differs between services.

Cost Optimisation: Understand the pricing models and how to reduce costs, including when to choose serverless over provisioned.

Data Movement: Know how data flows between services, when to use Data Factory, and how ETL and ELT differ.

Multi-Model Capabilities: Cosmos DB supports several data models. Know which one fits which scenario.

Recommended Study Resources

Use more than one type of resource:

Online learning platforms offer structured video courses. Microsoft Learn has free, official training modules mapped to the exam objectives.

Practice exams show your knowledge gaps. azureprep.com has DP-900 practice tests covering all domains, with a free preview before you subscribe. Repeating practice tests shows your improvement and builds stamina for a full exam.

Microsoft's official exam page has study materials and practice questions. The AZ-900 study guide covers foundation concepts that also apply to DP-900.

Hands-on work speeds up learning. Azure's free tier lets you create SQL databases, Cosmos DB instances and storage accounts, and even simple experiments make the concepts stick.

Exam Strategy and Tips

Read each question carefully and pick out the key requirements before looking at the answers. Eliminate the obviously wrong options first.

Watch for small differences between services. Some questions separate similar services on a specific capability or cost structure.

Allow roughly one minute per question. Flag difficult questions and come back to them if time permits.

Answer every question. A skipped question counts as incorrect, so a guess is better than no answer.

When you miss a practice question, read the explanation. Knowing why an answer is correct matters more than memorising it.

After Passing DP-900

DP-900 leads on to the advanced Azure data certifications. The AI-300 (Machine Learning Operations Engineer) and DP-700 (Fabric Data Engineer Associate) paths build directly on it.

Data Scientists typically take AI-300, which replaced DP-100 and covers machine learning operations with Azure Machine Learning.

Data Engineers now move on to DP-700, which replaced DP-203 (retired 31 March 2025) and covers building scalable data solutions on Microsoft Fabric.

Both paths need hands-on lab experience. Building Azure solutions in a free tier account develops the practical skills that DP-900 does not test.

Summary

DP-900 Azure Data Fundamentals certifies core data concepts and knowledge of Azure's data services. Most candidates need 2-4 weeks of focused study with Microsoft Learn modules and practice tests such as those on azureprep.com.

The exam rewards knowing the service categories, matching use cases to services and understanding basic data concepts. Practice exams show your weak areas before the real attempt.

The certification shows employers you have the data fundamentals, and it is the base for the advanced certifications. Booking the exam for no more than four weeks after you start studying keeps the plan on schedule.