Configuring discovery, access, and cost controls across a data lake

Managing a data lake on Google Cloud centers on using Dataplex to unify discovery, governance, and cost control across data stored in Cloud Storage and BigQuery. Dataplex organizes data into lakes, zones, and assets, applying consistent metadata, IAM-based access, and lifecycle policies without moving the underlying data. Effective management combines automated metadata harvesting, quality/profiling tasks, and storage class/lifecycle rules to balance accessibility with cost.

Must-know

• Dataplex logically organizes underlying Cloud Storage buckets and BigQuery datasets into lakes and zones (raw/curated), applying governance without duplicating or moving data.
• Data discovery is enabled via Dataplex's automatic metadata discovery, which scans assets and publishes tables into Dataplex catalog/BigQuery for querying, including schema and partition detection.
• Access control in Dataplex is managed through IAM roles at the lake, zone, or asset level, which propagate to the underlying Cloud Storage and BigQuery resources rather than replacing their native permissions.
• Cost control relies on Cloud Storage lifecycle management rules (transitioning objects to Nearline/Coldline/Archive or deleting them) and BigQuery table partition expiration/time-based partitioning to avoid unnecessary storage spend.
• Dataplex data quality and data profiling tasks run as managed Spark jobs, which incur separate compute costs beyond storage, so scheduling frequency should be tuned to actual governance needs.
• A common gotcha: granting IAM at the Dataplex lake/zone level does not automatically remove or override finer-grained IAM bindings already set directly on the Cloud Storage bucket or BigQuery dataset, so conflicting policies must be reconciled manually.

Coverage checked against the published exam guide on Aug 11, 2026.

These are independent practice questions, written against this certification's published exam guide. They are not the certification vendor's own questions, and not the real exam.