GCP Professional Data Engineer Practice Question 8: Sharing Datasets with Partners
Google Cloud Professional Data Engineer · Test 1 · question 8 of 10
A data provider must make a curated BigQuery dataset available to 40 partner organizations, each in its own Google Cloud organization. Partners must query the data from their own projects, the provider must not create per-partner copies, and the provider wants a catalog that partners can browse plus per-subscriber usage metrics, without granting partners IAM roles directly on the production dataset.
What should the provider do?
Show answer and explanation
Answer:
BigQuery sharing, the product formerly called Analytics Hub, is built for exactly this shape of problem. The provider creates a data exchange, publishes the dataset as a listing, and each subscriber gets a linked dataset in their own project: a read-only pointer to the source data, not a copy. Partners query it with their own compute, the exchange gives them a catalog to browse, subscriptions can be revoked individually, and the publisher gets subscriber-level usage metrics while the production dataset's IAM policy stays untouched.
Option A abandons the warehouse. Files behind signed URLs are stale snapshots, cannot be queried with SQL, and offer no per-partner revocation beyond waiting for a URL to expire.
Option B is the closest wrong answer: partners could technically query from their own projects. But it wires 40 external groups directly into the production dataset's IAM policy, provides no catalog or discovery surface, and yields no per-subscriber usage metrics, each of which the stem asks for.
Option C produces exactly what the stem forbids: 40 duplicated datasets, 40 times the storage cost, 40 refresh schedules to monitor, and data that is stale between copies.