PDE - Storing Data - Section 3.4

Design data platforms using Dataplex, Dataplex Catalog, BigQuery, and Cloud Storage, including federated governance models for distributed systems.

Design a data platform that uses Dataplex for unified governance across distributed storage and compute, BigLake for open-format federated queries, and Analytics Hub for cross-organisation dataset sharing, and define federated governance policies that work across domain boundaries.

DataplexBigLakeFederated governanceAnalytics Hub

Practice question for this objective

Free sampleStoring Datahard

A global media conglomerate is building a federated data platform across eight autonomous business units, each owning its own Google Cloud projects, BigQuery datasets, and Cloud Storage buckets of curated Parquet. The central data office must deliver two outcomes together: every unit must be able to publish governed, read-only data products to consumers in other Google Cloud organisations without duplicating storage or moving bytes, and the central office must maintain one searchable governance plane that classifies sensitive columns and applies common business metadata across all units' BigQuery and Cloud Storage assets. The platform team is selecting which TWO Google Cloud components, used together, deliver this federated governance and cross-organisation sharing model. Which TWO components should they choose? (Select TWO.)

  • AAnalytics Hub, so each business unit can publish data exchanges and listings that external subscribers query as linked datasets in their own projects, with the publisher revoking any single subscriber without affecting others. Correct
  • BDataplex with Dataplex Catalog, so the central office gets one searchable metadata plane that classifies sensitive columns with aspect types and applies common business metadata across BigQuery and Cloud Storage assets in every unit. Correct
  • CCloud Storage signed URLs distributed to each external organisation, so subscribers download the curated Parquet directly and the publishing unit revokes access by expiring the URLs.
  • DBigQuery authorised views shared by adding each external organisation's users directly to the underlying tables' IAM bindings, so consumers query the views in the publisher's project.
  • ECloud Data Loss Prevention inspection templates run on each bucket, so the central office tags sensitive columns and shares the de-identified output buckets with other organisations.
Federated governance across a distributed Google Cloud estate combines Analytics Hub for cross-organisation data product sharing with Dataplex and Dataplex Catalog for a unified, searchable governance plane. The two required outcomes map to two distinct services. Analytics Hub delivers zero-copy cross-organisation sharing because subscribers attach published listings as linked datasets and pay their own compute, while the publisher revokes subscribers individually. Dataplex with Dataplex Catalog delivers the central governance plane because aspect types and tag templates classify columns and attach business metadata across BigQuery and Cloud Storage assets in every project. Neither service replaces the other, so both are needed together.

Why A is correct: Analytics Hub is purpose-built for cross-organisation sharing: publishers expose listings that subscribers attach as linked datasets, query with their own slots, and have revoked individually, all without copying or egressing the source bytes.

Why B is correct: Dataplex and its Catalog provide the unified governance plane: aspect types and tag templates classify columns and attach business metadata consistently across BigQuery and Cloud Storage assets spanning many projects, satisfying the central searchable-classification outcome.

Why C is wrong: Signed URLs feel like a simple sharing mechanism, but they hand out raw object downloads that duplicate data into the subscriber's environment, bypass column-level governance, and give no searchable catalogue, so they meet neither outcome.

Why D is wrong: Authorised views are a real in-organisation pattern and sound plausible, but granting external users IAM on the publisher's tables forces the publisher to bear all query slots and egress and does not scale to cross-organisation sharing the way Analytics Hub listings do.

Why E is wrong: DLP inspection genuinely finds sensitive data and is tempting for the classification requirement, but it produces findings and de-identified copies rather than a federated searchable governance catalogue, and it provides no cross-organisation publishing surface, so it covers neither outcome on its own.

See more PDE practice questions, answers explained.

More in this domain

Back to all Storing Data objectives, or the PDE cert hub.

Examworthy is not affiliated with or endorsed by Google Cloud. Original, blueprint-aligned practice material only.