Amazon Neptune

Amazon Neptune is a fully managed graph database service on Amazon Web Services used to build and run applications that require connected datasets, storing graph data as vertices and edges.

AI Practitioner focus

  • Neptune is suited to highly connected data and knowledge graphs. Neptune Analytics adds vector similarity search so retrieval can combine semantic similarity with graph relationships.
  • The current FM objective lists Neptune as an embedding/vector option; Neptune ML is a separate graph-ML capability and is not the general name of the vector store.

Key points

  • The graph database engine stores billions of connections and queries them with millisecond latency.
  • Offers a choice of graph models and query languages:
    • Property Graph (PG) model, queried with the Apache TinkerPop Gremlin graph traversal language.
    • W3C standard Resource Description Framework (RDF) model, queried with SPARQL.
  • Highly available across three Availability Zones, with automatic failover to any of up to 15 low-latency read replicas.
  • Fault-tolerant storage by replicating two copies of data across three Availability Zones.
  • Continuous backup to Amazon S3 and point-in-time recovery from storage failures.
  • Automatically scales storage capacity and provides encryption at rest and in transit.
  • Common use cases: social networking, recommendation engines, fraud detection, and knowledge graphs — any workload built on highly connected data.

Sources