Skip to content

README.md

Latest commit

 

History

History
43 lines (30 loc) · 2.49 KB

README.md

File metadata and controls

43 lines (30 loc) · 2.49 KB

Google Spanner Graph

Globally distributed property graph on Cloud Spanner, combining ISO GQL graph queries with ScaNN-backed vector search in one storage layer.
Website cloud.google.com/spanner/docs/graph/overview
Model Label Property Graph
Status Production
License Commercial (commercial)
Query GQL (ISO/IEC 39075) / SQL
Vector strategy Integrated
HA Global, strongly consistent, multi-region
Deployment Google Cloud managed service

Overview

Spanner Graph is Google Cloud's native graph offering built on top of Cloud Spanner, Google's globally distributed, strongly consistent relational database. It exposes a property graph view over Spanner tables and allows querying that graph with ISO GQL alongside regular SQL.

Because the graph lives inside Spanner, it inherits Spanner's global consistency, multi-region replication, automatic sharding, and transactional guarantees. Spanner Graph is notable as the first hyperscaler graph offering built on the ISO GQL standard (ISO/IEC 39075).

Vector strategy — Integrated

Spanner exposes graph queries in ISO GQL alongside vector similarity functions (APPROX_COSINE_DISTANCE, APPROX_EUCLIDEAN_DISTANCE, APPROX_DOT_PRODUCT) and ScaNN-backed approximate nearest neighbor indexes in the same distributed SQL engine. Because graph tables and vector columns share the same storage layer, a single query can traverse a graph and perform vector search together without leaving the database.

Best fit

  • Applications that already depend on Spanner and want to add graph analytics without leaving the database
  • Globally distributed workloads that require strong consistency across regions
  • Teams wanting to adopt the ISO GQL standard rather than a vendor-specific graph query language
  • Workloads that need transactional graph and vector operations in one engine

Considerations

  • Pricing scales with Spanner's processing-unit and node model, which can be expensive for small or variable workloads
  • Only available as a managed Google Cloud service; no self-hosted deployment option
  • GQL tooling and ecosystem are still maturing relative to established graph query languages
  • Google Cloud lock-in by design

Sources