An Azure data platform is a collection of services that store, process, and manage data, from simple files to gigantic databases running across the globe. For the DP-300 exam, you need to know the different options so you can choose the right tool for the job, just like a builder chooses a hammer for nails and a saw for wood. This chapter gives you a high-level map of the Azure data landscape so you can understand where each service fits.
Jump to a section
A simple way to picture Overview of Azure Data Platform Options
A large apartment block's mailroom is the data platform for all the residents' communications. The mailroom itself is the physical building, the hardware and the space. Different residents have different needs. One resident receives only letters (structured data like a customer database). Another gets a daily delivery of heavy catalogues (semi-structured data like JSON files). A third receives giant boxes of building supplies (unstructured data like video files or documents). The mailroom manager must decide how to organise all this. One approach is to give each resident a dedicated, lockable mailbox (a relational database like Azure SQL Database — great for letters, secure, but inflexible if you want to store a box). Another approach is to use a large, open shelf where every package is just placed in order (a data lake like Azure Data Lake Storage — perfect for any size of box, but you need a system to find anything). A third approach is to use a specialised package-handling service that sorts parcels by content type (a NoSQL database like Azure Cosmos DB — fast for specific types of data, with different models for different packages). The mailroom is Azure's global infrastructure. The choices the manager makes about shelves, lockers, and sorting rules are the database services. The manager's job is to ensure each resident gets their mail efficiently and without loss, just as an Azure database administrator ensures data is stored, accessed, and protected correctly.
Microsoft Azure offers a huge range of data services. Think of them as a giant toolbox. To pass the DP-300 exam, you need to understand the major categories of tools in that box, not just each individual one. This is the 'Overview of Azure Data Platform Options'.
First, understand the fundamental model: cloud computing means you rent computing power and storage from Microsoft's data centres instead of buying and managing your own physical servers in your own building (on-premises). This gives you elasticity (you can scale up or down as needed) and pay-as-you-go pricing (you only pay for what you use).
The platform is built on several categories of services. The most important distinction for DP-300 is between relational databases and non-relational (NoSQL) databases.
Relational databases store data in tables with rows and columns, like a spreadsheet. They enforce a strict schema (defined structure) and use Structured Query Language (SQL) to query data. They are the classic choice for transaction processing (e.g., an e-commerce site's order table) because they guarantee ACID properties (Atomicity, Consistency, Isolation, Durability) — meaning every transaction is reliably processed. Key Azure relational services include Azure SQL Database (a fully managed version of Microsoft SQL Server), Azure SQL Managed Instance (more control, still managed), and Azure Database for PostgreSQL/MySQL/MariaDB (open-source engines).
Non-relational databases (often called NoSQL) do not rely on the table structure. They are designed for specific data models. Common types include key-value stores (like Azure Cosmos DB for NoSQL), document databases (also Cosmos DB, storing JSON documents), column-family stores (like Cassandra on Azure), and graph databases (like Azure Cosmos DB for Apache Gremlin). They are often preferred for massive scalability, flexible schemas (handling data that changes shape), and high-velocity data (like IoT sensor readings). They typically sacrifice some ACID guarantees for performance and scale.
Now, let's categorise the main Azure data services by purpose:
Transactional (OLTP) Databases: - Azure SQL Database: Fully managed, intelligent relational database. Great for line-of-business apps. - Azure SQL Managed Instance: Nearly 100% compatible with on-premises SQL Server, easier migration. - Azure Cosmos DB: Globally distributed, multi-model NoSQL. Supports multiple APIs (SQL, MongoDB, Cassandra, Gremlin, Table). Ideal for global apps needing low latency. - Azure Database for PostgreSQL/MySQL/MariaDB: Managed open-source relational databases.
Analytical (OLAP) Systems: - Azure Synapse Analytics: A limitless analytics service that combines big data and data warehousing. You can query both relational tables and data lakes using a single service. - Azure Databricks: A fast, collaborative Apache Spark-based analytics platform. Heavy on data science and machine learning.
Data Storage (Unstructured/Semi-structured): - Azure Blob Storage: Massively scalable object storage for any type of data (text, binary, images, videos). The foundation for data lakes. - Azure Data Lake Storage (Gen2): Built on Blob Storage, but with a hierarchical namespace for better file system performance. - Azure Files: Fully managed file shares in the cloud, accessible via SMB protocol (like a standard network drive).
Data Integration and Pipelines: - Azure Data Factory (ADF): A cloud-based ETL (Extract, Transform, Load) service that moves and transforms data.
Why this matters for DP-300: The exam tests your ability to recommend the right Azure data service for a given scenario. You will see scenarios like: 'A company needs a globally distributed database for a retail app with sub-millisecond latency. Which service?' Answer: Azure Cosmos DB. 'A company needs a fully managed relational database for an internal HR app in one region.' Answer: Azure SQL Database. The exam expects you to know the distinguishing features of each service, especially the trade-offs between SQL vs. NoSQL, managed vs. IaaS (Infrastructure as a Service), and transactional vs. analytical.
Key concepts also include high availability (e.g., zone-redundant configurations), disaster recovery (e.g., geo-replication), and security (authentication through Microsoft Entra ID, network security with firewalls). You do not need to memorise pricing details, but you must understand the architectural differences.
Identify the Data Type and Workload
First, determine if the data is structured (tables with rows/columns), semi-structured (JSON, XML), or unstructured (images, videos). Also determine if the workload is transactional (many small writes, ACID-compliant) or analytical (large aggregations, historical queries). This step narrows down the service category: relational vs. NoSQL vs. blob storage vs. data warehouse.
Evaluate Scalability and Distribution Needs
Ask: does the data need to be available globally with low latency? If yes, Azure Cosmos DB is a strong candidate because it supports multi-region replication with automatic failover. If the data stays in one region and has moderate growth, Azure SQL Database or Managed Instance may be sufficient. If massive scale (petabytes) is needed, consider Azure Data Lake Storage for raw data and Azure Synapse for analysis.
Determine the Level of Management Desired
Decide between PaaS (fully managed) and IaaS (self-managed). PaaS options (Azure SQL Database, Cosmos DB) handle backups, patching, and high availability automatically. IaaS options (SQL Server on a VM) give you full control over the operating system and database configuration, but you have to manage everything yourself. This choice impacts cost, time, and skills needed.
Consider Existing Technology Stack and Migration Path
If you are migrating from an on-premises SQL Server, evaluate the compatibility. Azure SQL Managed Instance offers the highest compatibility with features like SQL Agent, Service Broker, and cross-database queries. If you use open-source databases like PostgreSQL or MySQL, choose Azure Database for PostgreSQL or MySQL. This ensures minimal code changes.
Plan for Security and Compliance
Once the service is chosen, configure network security (firewalls, private endpoints), authentication (using Microsoft Entra ID instead of server logins), and encryption (Transparent Data Encryption). Also plan for audit logging and monitoring. For globally distributed data, understand data residency requirements (e.g., data must stay in the EU) to select the right Azure region.
Design for High Availability and Disaster Recovery
Configure built-in high availability features (zone-redundancy in Azure SQL Database, multi-region writes in Cosmos DB) and disaster recovery (geo-replication, failover groups). Test failover procedures. This step ensures your database remains available even during regional outages, which is a core responsibility of a database administrator.
Consider a medium-sized retail chain called 'NorthStar Books' that wants to modernise its IT systems by moving to Azure. As the database administrator (the person who designs and manages the databases), here is what you would actually do day-to-day regarding service selection.
The company has several data needs: 1. Customer Purchases Database: This powers their point-of-sale (POS) terminals at stores and the website's checkout. It must be ACID-compliant (guarantee every sale is recorded exactly once) and must support complex queries like 'How many books by a specific author did we sell last month?'. You would choose Azure SQL Database. You create a database, import the existing schema, set up a business-critical tier with zone-redundancy for high availability, and configure automated backups with a 7-day retention.
Product Catalogue with Images: The catalogue contains semi-structured data (JSON descriptions, prices, tags) and unstructured data (images, PDFs of author interviews). You need to store all this and serve it fast to the website. You would choose Azure Cosmos DB for the JSON product data (using the SQL API) and Azure Blob Storage for the images. You configure a CDN (Content Delivery Network) to cache images at edge locations globally.
Sales Analysis and Reporting: The marketing team wants to analyse historical sales trends over years. This is analytical (OLAP) work. You set up Azure Synapse Analytics to load transactional data from Azure SQL Database nightly. Using Synapse, you run complex queries that aggregate millions of rows. You also use Azure Data Factory to orchestrate the data movement: every night at 2 AM, a pipeline copies recent sales from Azure SQL Database, transforms the data (cleaning, aggregating), and loads it into Synapse.
Your daily tasks would involve:
Monitoring performance with Azure Monitor and SQL Insights.
Managing access: you grant specific Microsoft Entra ID (now Microsoft Entra ID) groups permissions via RBAC (Role-Based Access Control).
Optimising query performance by looking at slow queries and adding indexes.
Planning for disaster recovery: you set up geo-replication for Azure SQL Database to a secondary region in case of a regional outage.
Handling a data breach scenario: you would immediately review firewall rules, enable audit logging, and check for compromised credentials.
The key skill is knowing which service to pick and why, and then configuring it correctly for security, availability, and performance.
The DP-300 exam tests your understanding of Azure data platform options primarily through scenario-based questions. You will not be asked 'Define relational databases' directly. Instead, you will read a business description and pick the best service.
What they love to test:
Differentiating SQL vs. NoSQL: They will give a scenario with requirements for ACID transactions, complex joins, and a fixed schema (relational) vs. a scenario requiring flexible schema, global distribution, and high throughput (NoSQL). Watch for words like 'globally distributed', 'high throughput', 'sub-millisecond latency', 'flexible schema' which point to Azure Cosmos DB. Words like 'complex queries', 'joins', 'ACID transactions', 'business-critical transactional data' point to Azure SQL Database.
- Managed vs. Unmanaged (IaaS vs. PaaS): They love questions about moving an existing SQL Server database. Options include: - Azure SQL Database (PaaS) : Fully managed, you cannot control the underlying OS or SQL Server version. Best for new apps or simple migrations. - Azure SQL Managed Instance (PaaS with more control) : More compatible with on-premises SQL Server features (like SQL Agent, cross-database queries). Easier lift-and-shift. - SQL Server on Azure Virtual Machines (IaaS) : You manage everything (OS, SQL Server updates, backups). Gives most control, but most admin work. Best when you need OS-level access or have specific licensing requirements.
Data Warehousing vs. Transactional Database: They will ask you to choose between Azure Synapse Analytics (for analytical workloads) and Azure SQL Database (for transactional). Watch for keywords: 'reporting', 'aggregations', 'historical analysis', 'data warehouse' -> Synapse. 'Orders', 'inventory updates', 'customer inserts' -> Azure SQL Database.
Common Trap Patterns:
Azure Cosmos DB Trap: The exam might describe a scenario that *seems* like it needs a relational database (e.g., 'A retail store needs a database for transactions'), but then adds a requirement like 'and it must be globally distributed with multi-region writes'. Many beginners immediately think 'Azure SQL Database' for transactions, but Azure SQL Database does not natively do multi-region writes (you can do geo-replication but only one writer). The correct answer is Azure Cosmos DB because it supports multi-region write.
Azure Synapse vs. Azure Data Factory Trap: Some questions ask about moving and transforming data. Azure Data Factory is the ETL (Extract, Transform, Load) service to move data. Azure Synapse is the analytics engine to *query* the data. They are complementary. A question might ask 'Which service should you use to transform data from a CSV file and load it into a data warehouse?' The answer is Azure Data Factory, not Synapse.
Blob Storage vs. Azure Files Trap: If the requirement is to mount a file share via SMB protocol for multiple VMs (like a shared network drive), you use Azure Files. If you just need to store and serve large static files (images, videos), you use Azure Blob Storage.
Key Definitions to Memorise: - PaaS (Platform as a Service): Microsoft manages the infrastructure, you manage the database and data. - IaaS (Infrastructure as a Service): You manage everything (VMs, storage, networking). - OLTP (Online Transaction Processing): Many small transactions (e.g., inserting a sales order). - OLAP (Online Analytical Processing): Large complex queries over historical data (e.g., monthly sales report). - ACID: Guarantees for transaction processing. - Azure Cosmos DB: Globally distributed, multi-model NoSQL database. - Azure SQL Database: Fully managed relational database in the Microsoft SQL Server family.
Azure data services are categorised into relational databases (Azure SQL Database, Managed Instance, PostgreSQL/MySQL) and non-relational databases (Cosmos DB, Table Storage) based on the data model they support.
PaaS databases like Azure SQL Database and Cosmos DB reduce administrative overhead by handling backups, patching, and high availability automatically, while IaaS databases on VMs give you full control but require manual management.
Azure Cosmos DB is the only Azure database service that provides native global distribution with multi-region writes and single-digit millisecond latency, making it ideal for globally distributed applications.
Azure Synapse Analytics is designed for analytical workloads (OLAP) involving large-scale data aggregation and reporting, while Azure SQL Database is built for transactional workloads (OLTP) involving many small read/write operations.
Data movement and transformation are handled by Azure Data Factory (ETL service), not by the database or analytics engine itself, which means you often need multiple services working together in a data pipeline.
When migrating an existing SQL Server database to Azure, choose Azure SQL Managed Instance if you need near-complete feature compatibility, or Azure SQL Database if you are building a new application and can adapt to a slightly different feature set.
Role-Based Access Control (RBAC) using Microsoft Entra ID is the primary method for controlling access to Azure data services, overriding the older SQL Server login-based authentication in many modern deployments.
These come up on the exam all the time. Here's how to tell them apart.
Azure SQL Database (PaaS)
Fully managed by Microsoft: automatic backups, patching, replication.
Multi-tenant architecture, limited control over the underlying OS.
Built-in high availability and disaster recovery without manual configuration.
SQL Server on Azure VM (IaaS)
You have full control over the OS and SQL Server version, including all configurations.
You are responsible for all maintenance: backups, patching, performance tuning at the OS level.
Higher flexibility for customisations (e.g., third-party HA solutions, custom scripts).
Azure Cosmos DB (NoSQL)
Multi-model: supports document, graph, key-value, column-family.
Global distribution with multi-region writes and single-digit ms latency.
Elastic scalability of throughput and storage independently, with no schema enforcement.
Azure SQL Database (Relational)
Strict relational schema with tables, rows, columns, and relationships.
ACID-compliant transactions across multiple tables (e.g., bank transfers).
Complex query support including JOINs, aggregations, and subqueries within a single database.
Azure Synapse Analytics (OLAP)
A unified analytics platform for querying data warehouses and data lakes.
Data is stored and indexed within its own storage (dedicated SQL pool or serverless).
Used for high-performance analytical queries (aggregations, reporting, big data processing).
Azure Data Factory (ETL)
A cloud-based ETL service that orchestrates data movement and transformation.
Does not store data itself; it copies/transforms data between sources (e.g., Blob Storage, SQL Database) and destinations.
Used for building data pipelines: schedule jobs, transform using mapping data flows, and invoke activities (e.g., stored procedures).
Azure Blob Storage
Object storage accessed via HTTP/HTTPS (REST API).
Best for unstructured data: images, videos, backups, logs.
No file system semantics (no SMB protocol); cannot be mounted as a drive.
Azure Files
Fully managed file shares accessible via SMB protocol (like a standard network drive).
Can be mounted on multiple VMs simultaneously for shared access.
Ideal for legacy applications that expect a file system share or for lift-and-shift scenarios.
Mistake
Azure Cosmos DB is just a NoSQL database that only supports the document model.
Correct
Azure Cosmos DB is a multi-model database that supports SQL (document), MongoDB, Cassandra, Gremlin (graph), and Azure Table storage APIs. It is not limited to one data model.
Many beginners hear 'NoSQL' and assume it is one specific type (like only key-value or only document), but Cosmos DB is intentionally flexible.
Mistake
If you have a SQL Server database on-premises, the only way to migrate to the cloud is to lift-and-shift to a Virtual Machine (IaaS).
Correct
You can migrate to Azure SQL Managed Instance (PaaS) or even Azure SQL Database (PaaS) if you are willing to change some features like SQL Agent or cross-database queries. Managed Instance is designed for a near-zero-changes migration.
People think PaaS means losing all control, but Managed Instance preserves many features of full SQL Server while still being managed.
Mistake
Azure Synapse Analytics and Azure Data Factory do the same thing: they both move and analyse data.
Correct
Azure Data Factory is primarily a data integration and orchestration service that handles ETL/ELT (moving and transforming data). Azure Synapse Analytics is an analytics engine that stores and queries large amounts of data. They are often used together, but they are distinct services.
The names sound similar, and both deal with data pipelines, so beginners conflate them. The exam tests this distinction explicitly.
Mistake
You should always choose Azure Cosmos DB over a relational database because it is faster and more scalable.
Correct
Azure Cosmos DB is excellent for specific use cases (global distribution, flexible schema), but it does not support complex joins or ACID transactions across multiple documents in the same way a relational database does. For applications where data consistency and complex queries are critical, a relational database like Azure SQL Database is often a better choice.
Beginners see the word 'globally distributed' and think it is always the best option, not understanding the trade-offs.
Mistake
Azure Blob Storage and Azure Files are basically the same thing — they both store files.
Correct
While both store files, they have different protocols and use cases. Azure Blob Storage is accessed via HTTP/HTTPS (REST API) and is optimised for large-scale object storage (images, backups). Azure Files is accessed via SMB protocol (like a standard network share) and is designed for applications that need a mounted file system, such as legacy apps running on VMs.
Both start with 'Azure' and involve 'storage', so people assume they are interchangeable. The exam tests scenario-based selection based on access protocol.
Reveal each answer, then mark whether you got it right. Score 60%+ to unlock the next chapter.
Azure SQL Database is a fully managed, multi-tenant PaaS database with a more limited set of features (e.g., no SQL Agent, no cross-database queries). Azure SQL Managed Instance provides near-100% compatibility with on-premises SQL Server, including SQL Agent, Service Broker, and cross-database queries, while still being fully managed.
Use Azure Cosmos DB when you need global distribution with multi-region writes, support for multiple data models (document, graph, key-value), and single-digit millisecond latency regardless of scale. Use Azure SQL Database when you need ACID transactions, complex joins, a fixed relational schema, and you do not need global distribution.
Azure Synapse Analytics is an analytics engine for storing and querying large amounts of data (a data warehouse and big data service). Azure Data Factory is an orchestration and ETL service that moves and transforms data between sources and destinations. They are often used together: Data Factory moves data into Synapse for analysis.
Yes, you can use your existing SQL Server licenses with Software Assurance to get a reduced rate on Azure SQL Database, Managed Instance, or SQL Server on Azure VMs. This is called Azure Hybrid Benefit. It can significantly lower costs.
Azure Blob Storage is a general-purpose object storage service for any type of data (files, backups, images). Azure Data Lake Storage Gen2 builds on Blob Storage by adding a hierarchical namespace, which makes it more like a file system and improves performance for big data analytics workloads (e.g., with Spark or HDFS).
You should use private endpoints or service endpoints to connect over the Microsoft backbone network (not the public internet). Enable firewall rules to limit IP addresses. Use Microsoft Entra ID authentication instead of SQL Server logins, and enable Transparent Data Encryption for data at rest. For client connections, use encrypted connections (TLS 1.2+).
You've finished Overview of Azure Data Platform Options. Continue through the DP-300 study guide to build a complete picture of the exam.
Done with this chapter?