Azure Cosmos DB

Azure Cosmos DB
Name	Azure Cosmos DB
Developer	Microsoft
Released	10 May 2017
Operating system	Azure cloud platform
Genre	NoSQL, Multi-model database
License	Proprietary

Azure Cosmos DB. It is a globally distributed, multi-model database service designed and operated by Microsoft on its Azure cloud platform. Launched in 2017, it provides turn-key global distribution, elastic scaling of throughput and storage, and guarantees for latency, throughput, consistency, and high availability. The service is engineered to support mission-critical applications at a global scale, offering comprehensive service level agreements.

Overview

Azure Cosmos DB was developed by Microsoft as a foundational component of its Azure cloud strategy, building upon earlier projects like DocumentDB. It is architected as a schema-agnostic database system that natively supports multiple data models, including document, key-value, graph, and column-family, accessed via various APIs. The core engine is built with global distribution and horizontal scale as first-class principles, utilizing a resource governance system called Request Units for predictable performance. It is deeply integrated with other Azure services like Azure Functions, Azure Synapse Analytics, and Azure Machine Learning.

Data model

The fundamental data model is based on an atom-record-sequence structure, where items are stored in containers within a database account. Containers are the units of scalability for both throughput and storage, partitioning data across logical partitions using a user-defined partition key. Data is represented as schema-free JSON items, which can be projected as documents in the SQL API, nodes and edges in the Gremlin API, or rows in the Cassandra API. This flexibility allows the same underlying storage to support applications traditionally built on MongoDB, Apache Cassandra, or Azure Table Storage.

Consistency levels

The service offers five well-defined consistency models, allowing developers to make precise trade-offs between performance and data currency. These levels, from strongest to weakest, are strong, bounded staleness, session, consistent prefix, and eventual consistency. This spectrum is a core feature, providing more granularity than the typical binary choice offered by systems like Amazon DynamoDB or Google Cloud Spanner. The consistency level can be set at the database account level and overridden at the individual request level for fine-grained control.

APIs and protocols

Azure Cosmos DB provides multiple wire protocol-compatible APIs, enabling use with existing toolkits and drivers. The primary API is a core SQL dialect for querying JSON items. For NoSQL workloads, it offers the MongoDB wire protocol, the Apache Cassandra Query Language, and the Gremlin graph traversal language. It also supports a key-value API via Azure Table Storage and a RESTful HTTP API for serverless scenarios using Azure Functions. This multi-model approach differentiates it from more singular services like Amazon DynamoDB.

Global distribution

A defining capability is turn-key global distribution, allowing data to be replicated to any number of Azure regions worldwide with a single click. The replication system uses a multi-master architecture where every region can accept both reads and writes. This supports low-latency access for globally dispersed users and is foundational for achieving comprehensive disaster recovery and business continuity plans. The distribution is managed via Microsoft's global network backbone, which connects its massive data center infrastructure.

Use cases

Typical use cases include globally scalable web applications, mobile applications, and Internet of Things platforms that require single-digit millisecond response times. It is heavily utilized in gaming for player profile and state management, in retail for catalogs and shopping carts, and in social media for activity feeds. Its multi-model nature makes it suitable for modern microservices architectures where different services, such as those built for Netflix or eBay, might prefer different data paradigms while sharing a unified backend.

Category:Cloud databases Category:Microsoft Azure Category:NoSQL