CONCEPT Cited by 1 source

Multi-tenant graph platform

A multi-tenant graph platform is a shared infrastructure where multiple graph use cases (tenants) run on the same engine with namespace isolation, centralized management (schema enforcement, index lifecycle, API generation), and shared operational tooling.

Motivation

Before Airbnb built their centralized platform, graph adoption was fragmented across four anti-patterns:

1. Relational "graphs" — modeling nodes/edges in SQL → expensive joins during traversal.
2. Offline graphs — building in the data warehouse → daily-stale snapshots, no real-time serving.
3. DIY open source — self-managing community graph DBs → high operational toil per team.
4. Managed PaaS — third-party vendors → lock-in, performance ceilings, opaque internals.

A centralized platform eliminates per-team ops burden while providing a consistent, tunable, observable substrate.

Design properties

• Namespace isolation — each tenant's data is logically separated.
• Schema enforcement — the platform validates graph schemas per tenant.
• Index management — centralized control over property indexes.
• Schematized APIs — auto-generated typed interfaces (e.g., Thrift) per tenant's schema.
• Shared observability — distributed tracing, metrics, alerting across all tenants.

Seen in

• sources/2026-05-19-airbnb-scaling-identity-graph-unified-knowledge-graph-infrastructure — Airbnb's knowledge graph infrastructure is a multi-tenant platform on JanusGraph + DynamoDB. Tenants include identity graph (7B nodes), inventory knowledge graph, fraud detection, and data lineage — each in an isolated namespace with schema enforcement and managed Thrift APIs.

Related

• systems/airbnb-knowledge-graph-infrastructure — canonical instance
• patterns/tenant-per-application — related isolation pattern
• concepts/knowledge-graph — the data model tenants typically serve