Databases Across the Three Clouds

Relational, NoSQL, warehouse — each of these has a named service on every major cloud. The names differ, the underlying concepts do not. This topic maps the database catalog so that when you see "DynamoDB," "BigQuery," or "Cosmos DB" in a job description or an architecture diagram, you know immediately what family it belongs to.

There is also an important distinction to understand here: some database services run open-source engines (PostgreSQL, MySQL, Redis) that you can run anywhere, while others are proprietary cloud inventions that only exist on one provider. That difference has real consequences for how freely you can move between clouds.

The Mapping: Same Families, Different Names

All three clouds cover the full database landscape: managed relational, managed NoSQL, data warehouse, and managed cache. Each is a different cloud brand for the same fundamental kind of store. The services in each row below do the same job.

Signature Offerings Worth Knowing

Two services stand out as genuinely distinctive. Google's BigQuery was one of the first truly serverless, massively-parallel data warehouses — you query petabytes without provisioning any servers, and you pay per query rather than per running cluster. AWS's Redshift and Azure Synapse Analytics are comparable, but BigQuery's fully serverless design was a step ahead when it launched and remains a benchmark. It is worth knowing by name.

Azure's Cosmos DB is also unusual. Where most NoSQL databases specialize in one model (DynamoDB is key-value and document; Firestore is document), Cosmos DB is multi-model — the same service can behave as a document store, a key-value store, a wide-column store, or a graph database, depending on the API you use. That breadth is architecturally distinctive; no other cloud has a direct equivalent.

Open vs Proprietary: Why It Matters

Some managed database services run open-source engines: PostgreSQL, MySQL, MariaDB, Redis. Because these engines are open standards, you could move your database from AWS RDS to Google Cloud SQL to Azure Database without rewriting any queries — the SQL dialect, the behavior, and the backup format are the same. Open-source managed databases give you portability: the cloud runs the engine, but the engine is not owned by the cloud.

Other services are proprietary: DynamoDB (AWS), BigQuery (Google), Cosmos DB (Azure), and Google's Spanner, for example, are cloud-only inventions. They may be excellent tools for the right problem — BigQuery's scale is genuinely hard to replicate elsewhere — but they create lock-in. If you build a large application on DynamoDB's specific API, migrating to a different cloud means rewriting significant parts of your data layer. The decision to use a proprietary database is a real architectural choice, not just a service-name swap.

A practical rule of thumb: open-source managed databases (RDS for PostgreSQL, Cloud SQL, Azure Database) are portable; proprietary managed databases (DynamoDB, BigQuery, Cosmos DB) are not. Both have their place; the difference is worth understanding before you commit.

Reading the Menu

In practice, recognizing a database service in the wild is mostly pattern-matching. Services with "RDS," "SQL," or a database name (PostgreSQL, MySQL) in them are relational. "DynamoDB," "Firestore," "Cosmos DB" are NoSQL. "Redshift," "BigQuery," "Synapse" are warehouses. "ElastiCache," "Memorystore," "Cache for Redis" are caches. The categories from this chapter are the decoder ring for the whole catalog.