NotesSystem

Neo4j

Also: Neo4j Graph Database · Neo4j Aura

A native property-graph database queried with Cypher. Its defining choice is index-free adjacency: each node's storage leads straight to its relationships, so a traversal hops without a global index lookup.

Also in: Data Models & Query Languages › In the real world

Design choices

  • Graph Model: labelled property graph: nodes with labels and properties, typed directed relationships with properties.
  • Native storage: the older record formats chain each node's relationships as linked records; the block format (Enterprise default since 5.22) keeps a node's properties and relationships in its own block, with dense nodes overflowing into a tree sorted by type and direction. Either way, traversal cost tracks the part of the graph visited, not its total size.
  • Declarative Query Language: Cypher, pattern syntax like (a)-[:KNOWS*1..3]-(b). Cypher was the main basis of GQL (ISO/IEC 39075, 2024). Since 2025 Cypher is versioned separately from the server (Cypher 5 frozen, Cypher 25 evolving), and new releases add GQL conformance features.
  • Transactions: ACID, with clustered deployments replicating through Raft.
  • Versioning: calendar versions for the server since 2025.01.

Notable trade-offs

  • Excellent at deep, variable-length traversals; less suited to bulk scans and aggregates over the whole data set.
  • A highly connected graph is hard to shard, because traversals cross partition boundaries.

Used by

  • ICIJ (2016): the Panama Papers' 11.5 million leaked documents became a Neo4j graph of people, companies and intermediaries, explored with Linkurious; reused for the Paradise Papers (2017).