Cosmos DB, Microsoft’s managed database cloud service, received several updates today aimed at making it useful for a wider variety of users. One key addition is a preview of support for using the Cassandra NoSQL database API to run operations inside the system.
This provides another tool for categorizing and analyzing data inside Cosmos DB, which already supports Gremlin, MongoDB, and SQL. Microsoft also announced the general availability of an Apache Spark Connector that allows businesses to run real-time data analysis tasks over data stored in Cosmos DB.
In addition, Microsoft is increasing the strength of the guarantees it makes for Cosmos DB’s availability. The company said that it will ensure 99.999 percent read availability for databases stored with Cosmos across multiple regions. That’s up from a 99.99 percent guarantee and comes alongside a trio of other guarantees around latency, throughput, and consistency.
Cosmos DB is designed to provide customers with a globally distributed database that they can use to power applications without having to manage the complexity of maintaining copies of data in multiple disparate locations.
Fully managed, globally distributed databases are services that many cloud providers leaned into as a key tool for attracting business customers. They highlight some of the key promises of the cloud, with support for automatic scalability and the ability to reduce the workload placed on operations engineers.
The news comes a day after Google announced the general availability of its Cloud Spanner database service, which also guarantees 99.999 percent uptime.
While the two systems are both fully managed databases with global availability and multi-region replication at their core, they approach the problem in different ways. Cosmos DB is oriented around a NoSQL approach, while Cloud Spanner is built to behave like a traditional relational database.
One of Cosmos DB’s key features is the ability for users to select their preferred consistency model. That’s based on CAP Theorem, which argues that databases can guarantee two of three traits: consistency, availability, and partition tolerance. If developers don’t need perfect consistency, they can get additional benefits when it comes to availability and performance, and Cosmos DB provides them with the tools to make those choices.
According to Azure CTO Mark Russinovich, the most popular consistency model is Session Consistency, which guarantees that a particular user of a database application will have all of their reads and writes be internally consistent, though data from other users may not match the same criteria.
This news is part of a raft of other announcements that Microsoft released as part of its Connect conference in New York today. The company revealed that it’s joining the open source MariaDB Foundation to support the development of that database software, as well as its use on Azure.
Microsoft also launched an Azure Databricks service that’s designed to make it easier for developers and data scientists to collaborate on real-time analytics using a cloud-hosted platform based on the Apache Spark project. That service also features native integration with Cosmos DB.
Correction 7:50 a.m. Pacific: This article previously said that the third trait in CAP Theorem is performance. It is not. The third trait is partition tolerance.