ArangoDB — Native Multi-Model Database (Documents, Graphs, Key-Value)

Introduction

ArangoDB lets you model documents, graphs, and key-value pairs in the same database — and query them all with AQL (ArangoDB Query Language), which reads like SQL for graphs. Instead of running MongoDB + Neo4j + Redis, one cluster handles all three workloads.

With over 14,000 GitHub stars, ArangoDB is used by Oxford University Press, Cisco, FlightAware, and many others. It scales horizontally via automatic sharding (SmartGraphs, SmartJoins) and supports ACID transactions across documents and graph edges.

What ArangoDB Does

ArangoDB stores JSON documents in collections (like MongoDB). Edge collections link documents to form graphs. Every document also has a key, so it doubles as a key-value store. AQL queries combine document lookups, joins, and graph traversals in one pass.

Architecture Overview

Clients (HTTP / Foxx / Drivers)
        |
  [ArangoDB Coordinator]
   AQL parser, planner, router
        |
  +------+------+------+
  |      |      |      |
[DB Server] [DB Server] [DB Server]
  RocksDB engine
  Shards of docs + edges
        |
  [Agency (Raft-based config store)]
   cluster metadata + leader election

Data models on one engine:
  documents  -> JSON collections
  graphs     -> edge collections, AQL traversals
  key-value  -> document _key lookups
  search     -> ArangoSearch views (inverted index)

Self-Hosting & Configuration

// Graph traversal: find friends-of-friends under 35
FOR v, e, p IN 2..2 OUTBOUND "people/alice" knows
  FILTER v.age < 35
  RETURN DISTINCT v.name

// Geo + document + graph in one query
FOR place IN GEO_POINT(places, [-73.99, 40.73], 0, 1000)
  FOR visitor, visit IN 1..1 INBOUND place visited
    FILTER visit.rating >= 4
    RETURN { place: place.name, visitor: visitor.name }

// ArangoSearch (full-text) across many collections
FOR doc IN combined_view
  SEARCH ANALYZER(
    doc.title IN TOKENS("distributed database", "text_en")
    OR doc.description IN TOKENS("cluster", "text_en"),
    "text_en")
  SORT BM25(doc) DESC
  LIMIT 20
  RETURN doc

Key Features

• Multi-model — documents, graphs, key-value, search in one DB
• AQL — unified query language for all data models
• ACID transactions — cross-collection, cluster-wide
• SmartGraphs — sharded graphs that preserve locality
• SmartJoins — co-located joins across sharded collections
• ArangoSearch — full-text, BM25 + tf-idf, multi-language analyzers
• Foxx — server-side JavaScript microservices inside the DB
• Horizontal scale — automatic sharding, replication, failover

Comparison with Similar Tools

FAQ

Q: ArangoDB vs MongoDB + Neo4j? A: If your app needs both documents and graphs, one database is simpler than two. ArangoDB handles both with similar performance to specialists for most workloads.

Q: Is ArangoDB open source? A: Yes, Apache-2.0. There is an Enterprise Edition (paid) with SmartGraphs extras, hot backups, audit log, and more.

Q: Performance vs Neo4j for deep traversals? A: Neo4j is optimized for pure graph OLTP and often faster for deep traversals on small datasets. ArangoDB is competitive and often wins on mixed document + graph workloads.

Q: What is Foxx? A: Foxx is ArangoDB's embedded JavaScript microservice framework. Deploy REST APIs that run inside the DB process, eliminating round-trip latency for complex business logic.

Sources

• GitHub: https://github.com/arangodb/arangodb
• Docs: https://docs.arangodb.com
• Company: ArangoDB Inc.
• License: Apache-2.0