Netflix faces challenges with duplicated and inconsistent data models, inconsistent terminology, data quality issues, and limited data connectivity across its systems.
To address these challenges, Netflix introduces UDA (Unified Data Architecture) to define a model conceptually once and reuse it consistently across different systems.
UDA allows for registering and connecting domain models, cataloging and mapping domain models to data containers, transpiling domain models into schema definition languages, and moving data between data containers.
UDA serves as a knowledge graph platform that connects domain models to data containers through mappings and a metamodel called Upper, enabling semantic interoperability and consistent data representation.
Netflix utilizes UDA in production for systems like Primary Data Management (PDM) and Sphere, which manage controlled vocabularies and operational reporting, respectively, by leveraging UDA's capabilities for consistent data representation and integration.
UDA adopts a named-graph-first information model to ensure resolution, modularity, and governance across the entire graph, addressing challenges with existing ontology frameworks.
Netflix uses Upper as a domain modeling language in UDA, enabling formal descriptions of business and system domains, paving the way for connected data semantics and interoperability.
Projections in UDA produce concrete data containers like GraphQL and Avro schemas, ensuring consistent public contracts across systems and enabling automated data handling.
Overall, UDA revolutionizes Netflix's data modeling approach, providing a unified knowledge graph for consistent, connected, and discoverable data across the organization.
Future applications of UDA include supporting additional projections like Protobuf/gRPC and materializing instance data knowledge graphs for querying and profiling.
UDA represents a significant step in enhancing data management at Netflix, addressing data consistency, integration, and interoperability challenges within its complex ecosystem.