Microsoft debuts Azure Cosmos DB, a globally distributed database with five consistency choices, and new MySQL and PostgreSQL options for Azure Database
Arguably the biggest Azure news to come out of Microsoft's Build 2017 developer conference is the debut of Azure Cosmos DB.
Context & Ripple Effects
The Cosmos DB launch caps a two-year buildout of Azure's data layer: after bringing SQL Server into Linux containers via Docker Datacenter in 2016, Microsoft spent Build 2017 making its database story global-first, debuting Cosmos DB as a globally distributed store whose five selectable consistency levels let developers trade latency against correctness per request. Alongside it, managed MySQL and PostgreSQL options extend Azure Database beyond SQL Server toward the open-source stacks enterprises actually run.
The follow-through came fast: within months Microsoft was joining the MariaDB Foundation to add another fully managed open-source engine, and in 2019 it bought Citus Data to turn PostgreSQL itself into a distributed database — suggesting Cosmos DB's distribution thesis became the template for the whole portfolio.
First-order effects
- Azure developers gain a single API surface over a globally replicated database with five tunable consistency choices, removing the need to hand-build multi-region replication for planet-scale apps.
- Teams standardized on MySQL and PostgreSQL get first-party managed hosting on Azure Database, lowering the switching cost of moving open-source workloads off self-run servers or rival clouds.
Second-order effects
- Competing clouds must answer a multi-model, multi-consistency distributed database with their own managed offerings — a contest that culminated seven years later in AWS's Aurora DSQL, which pairs strong consistency with PostgreSQL compatibility.
- Open-source database projects gain leverage as hyperscalers compete to host them: Microsoft's MariaDB Foundation membership and Citus acquisition show vendors buying into communities rather than building purely proprietary alternatives.
Third-order effects
- Distributed-by-default becomes the baseline expectation for cloud databases, with hyperscalers differentiating on how they reconcile global scale with familiar interfaces like PostgreSQL rather than on raw managed hosting.
- Consistency-as-a-configuration — letting applications pick among five isolation guarantees instead of one vendor default — points toward databases marketed as SLA-tunable infrastructure, a shift regulators and enterprise buyers may eventually scrutinize as workloads concentrate across regions.
The trend: Cloud databases are converging on globally distributed architectures wrapped in open-source-compatible APIs, with each hyperscaler acquisition and launch — Citus, MariaDB hosting, Aurora DSQL — reinforcing the same playbook.