Governing data consistently across a federated, distributed system

A federated governance model lets independent domain teams own and manage their data products while adhering to centrally defined policies for security, quality, and metadata standards. On Google Cloud, this is implemented using Dataplex to organize data into logical domains (lakes and zones) while enforcing consistent governance, cataloging, and access controls across BigQuery, Cloud Storage, and other sources. This approach balances domain-level autonomy with organization-wide consistency, which is central to modern data mesh architectures.

1 · Learn the must-know

• 1Dataplex provides the unifying layer for federated governance by letting you organize distributed data (BigQuery datasets, Cloud Storage buckets) into logical lakes and zones without physically moving the data.
• 2Dataplex automatically discovers and catalogs metadata from registered assets, making data discoverable and queryable through a unified metadata layer (Dataplex Catalog) regardless of which team owns it.
• 3Centralized policies (data quality rules, IAM permissions, data classification via sensitive data protection) can be defined once at the lake or zone level and automatically inherited by underlying assets, ensuring consistency across domains.
• 4IAM and Dataplex's fine-grained access controls allow domain teams to retain administrative control over their own data (zones/assets) while central governance teams enforce organization-wide security and compliance policies.
• 5Data quality tasks and profiling in Dataplex can be scheduled per zone or asset, allowing each domain to define quality rules relevant to their data while reporting up to a central governance dashboard.
• 6A common gotcha: federated governance requires careful organization of resource hierarchy (organization, folders, projects) and consistent tagging/classification taxonomy upfront, or metadata and policies become fragmented across domains, undermining the model's benefits.

2 · Check your understanding