Tencent Oula Data Governance Platform: Architecture, Practices, and Solutions

Tencent Oula Platform is a DataOps‑based, one‑stop data governance solution launched by Tencent PCG in 2019, covering the full lifecycle of data production, management, governance, and usage, and providing four core capabilities: data discovery, asset factory, metric platform, and governance engine.

Platform Construction Goals

Enable industrialized data production and application.

Support two typical scenarios: fixed business processes/data models and complex, rapidly changing business needs requiring multi‑technology solutions.

Provide three core abilities: efficient business‑process customization, robust data‑model management, and unified compute‑storage services.

Overall Architecture

The platform combines platform capabilities with governance projects to drive best‑practice implementation, emphasizing data standards, full‑link metadata, unified data entities/models/services, and a governance evaluation system.

Data Development Governance

Using DataOps principles, Oula addresses common warehouse issues such as inconsistent calculations, data redundancy, and unclear metric sources by promoting standardized modeling, visual design, and a unified metric store (tMetric) to ensure consistent metric definitions across reporting, analysis, and experimentation platforms.

The modeling layer distinguishes physical, logical, and conceptual models, while DataOps orchestrates the entire development pipeline—from requirement gathering, design, testing, to release and monitoring.

Unified Metric Store (tMetric)

tMetric consolidates metric definitions, provides standardized modeling, enforces metric consistency, offers certification, and builds an ecosystem for cross‑platform metric reuse. Metrics are classified as atomic or derived, and can be accessed via API or materialized into DWS/ADS tables.

Data Map and Service Construction

To avoid data silos, Oula builds a Data Fabric based on metadata, enabling users to discover, understand, and request data through a unified data service layer that automatically provisions APIs for Redis, ClickHouse, MySQL, etc., with built‑in monitoring and auto‑scaling.

Q&A Highlights

Asset health scoring evaluates five dimensions—standard, security, quality, cost, and application—using metadata‑driven rules. Data lineage is captured via Open API metadata reporting, linking upstream ingestion to downstream reports, though storage of lineage itself typically relies on mature graph databases.

The session concludes with a summary of Oula's end‑to‑end governance loop, from metadata collection to quality evaluation and asset health improvement.