A managed relational database service runs a familiar database engine - MySQL, PostgreSQL, SQL Server, Oracle - but hands the operational toil to the cloud provider. You still get the same SQL, the same drivers, and the same transactional guarantees; what you give up is patching, backups, replication setup, failover plumbing, and hardware. Amazon's RDS (Relational Database Service) is the best-known of these, and "RDS" has become shorthand for the whole category of managed relational databases across clouds.
What "managed" actually covers
The value is in the work you no longer do. A managed service handles automated backups and point-in-time recovery, minor-version patching, monitoring and metrics, and - with a click - high availability through a standby replica in another availability zone that takes over automatically on failure. It also makes read replicas and storage scaling routine operations rather than projects. You manage the schema, the queries, the indexes, and the data; the provider manages the machine the database runs on.
The engines
Managed relational services wrap both open-source and commercial engines. The open-source options are the cheapest and most portable: MySQL and PostgreSQL, run as managed instances. The commercial engines - Microsoft SQL Server and Oracle Database - let you run enterprise databases without managing the underlying servers, though licensing follows you. And there are cloud-native engines built specifically for the managed model: Amazon Aurora reimplements the storage layer beneath MySQL and PostgreSQL for better performance and durability. Across clouds the same idea appears under different names - Google Cloud SQL and Azure Database are the GCP and Azure equivalents.
OLTP, not analytics
Managed relational databases are built for transactional (OLTP) workloads - the reads and writes that power applications, where you fetch and update individual rows fast and need ACID guarantees. They're not analytical warehouses: large aggregations over the whole table will run, but slowly, and they'll compete with the application traffic. The common architecture pairs a managed relational database as the operational system of record with a separate cloud data warehouse for analytics, fed by a pipeline that moves data between them.
Choosing one
- New application, cost-sensitive, no engine lock-in? Managed PostgreSQL or MySQL.
- Want the highest performance and durability on a MySQL/Postgres API? Aurora.
- Existing SQL Server or Oracle estate? The managed version of that engine, to lift-and-shift without re-platforming.
- Already standardized on a cloud? Its native service (RDS on AWS, Cloud SQL on GCP, Azure Database) for the tightest integration.
The decision is mostly about engine compatibility and cloud fit; the operational benefits of "managed" are broadly similar whichever you pick.