
Distributing modules, not just servers—Airbnb's unique architectural take on GraphQL scaling.
Airbnb has transitioned Viaduct, its core data service mesh, to a community-driven open-source project with the 1.0 release. It introduces a 'module-based multi-tenancy' approach that solves the overhead of server-heavy federated architectures in massive engineering orgs.
Highly recommended for platform engineers building company-wide data APIs or organizations that have outgrown traditional single-service GraphQL setups.
Large organizations face challenges managing a central GraphQL schema as it scales, struggling to balance developer autonomy across many teams with the need for stable governance and a unified interface.
Airbnb developed Viaduct, a multi-tenant runtime that distributes development through modules rather than servers, allowing teams to contribute SDL and resolvers to a shared graph without managing infrastructure.
The 1.0 release establishes a stable public API and Kotlin binary compatibility, evolving Viaduct into a community-driven project where hundreds of teams contribute to a coherent data layer efficiently.
Trade-off
Adopting a shared runtime creates risks where one module's failure could impact others, necessitating investment in advanced sharding and probabilistic testing to maintain isolation and stability.
A system that provides a single GraphQL-based interface for accessing and interacting with any data source, abstracting complexity from clients.
A shared execution environment that hosts independently developed tenant modules owning specific portions of the schema.
An architectural approach that composes multiple subgraphs into a single unified graph through a router.




