Multi-Model Database

The previous lesson made the case for polyglot persistence, different databases for different data patterns. But you also saw the costs: operational complexity, data synchronization headaches, and the need for a team large enough to manage six database technologies.

What if a single database could handle relational data, documents, graphs, key-value lookups, and search, all in one engine? You'd get the right data model for each use case without the operational overhead of multiple databases.

This is the promise of multi-model databases. Instead of PostgreSQL for relations, MongoDB for documents, Neo4j for graphs, and Redis for key-value, you deploy one database that supports all of these models natively.

ArangoDB, Azure Cosmos DB, SurrealDB, and FaunaDB all claim this ground. The question is whether the promise matches reality.

A multi-model database provides multiple data model interfaces on top of a single storage engine.

ArangoDB

Supports documents, graphs, and key-value in a single database with a unified query language (AQL).

// Document query
FOR product IN products
 FILTER product.category == "electronics" AND product.price < 500
 RETURN product

// Graph traversal (same database, same query language)
FOR v, e, p IN 1..3 OUTBOUND "users/alice" follows
 RETURN v.name

// Key-value lookup
DOCUMENT("users/alice")

One database, one query language, one deployment. The storage engine underneath handles all three models.

Azure Cosmos DB

Microsoft's globally distributed multi-model database. Supports five APIs:

• SQL API (document store)
• MongoDB API (wire-compatible with MongoDB)
• Cassandra API (wire-compatible with Cassandra)
• Gremlin API (graph traversal)
• Table API (key-value)

Same underlying storage engine, different API facades. You choose the API when creating a container.

SurrealDB

Newer entrant. Combines document, graph, and relational models with a SQL-like language. Designed for real-time applications with built-in authentication and permissions.