Skip to main content
Definition Website Computer and Technology

Graph Database

Subject: Computer and Technology noun /ˈɡræf ˈdeɪ.təˌbeɪs/ 1 definition

  1. Definition (noun)

    Entry #140 · Updated

    A Graph Database is a specialized NoSQL database management system that stores, manages, and queries data as interconnected graph structures composed of nodes (entities), edges (relationships), and properties (attributes). Rather than resolving expensive multi-table join operations at query time, it persists direct connections between adjacent records—often using index-free adjacency—so queries can traverse deep relationship paths in near-constant time per hop regardless of overall dataset size. While relational and document databases optimize tabular aggregation or isolated record retrieval, graph databases excel at topology-driven workloads such as fraud detection, recommendation engines, knowledge graphs, and dependency analysis.

    Common Types

    • Labeled Property Graph (LPG) Databases: Model entities as labeled nodes and relationships as directed edges that can both hold key-value properties, optimizing fast operational traversals with query languages such as Cypher, GQL, or Gremlin.

    • RDF Triple Stores: Store facts as standardized Resource Description Framework (RDF) subject-predicate-object triples identified by Uniform Resource Identifiers (URIs), supporting formal ontologies, semantic reasoning, and linked data queries via SPARQL.

    • Hypergraph Databases: Generalize standard pairwise edges into hyperedges that connect arbitrary sets of nodes at once, enabling native modeling of complex many-to-many and n-ary associations without auxiliary intermediate nodes.

    • Multi-Model Graph Databases: Layer graph storage and traversal APIs alongside document, key-value, or relational engines in one platform, trading some native pointer-chasing speed for unified querying across mixed data models.

    Example

    During online checkout, a payment platform queries a property graph database linking Account, Card, Device, and IPAddress nodes through USED_BY and SHARED_WITH edges. A single three-hop traversal identifies within milliseconds that four newly registered accounts share the same device fingerprint and billing address as a flagged fraud ring, blocking the transaction before authorization completes.