Free GCP-PDE sample questions

Correct answers: Configure Customer-Managed Encryption Keys (CMEK) using Cloud KMS for the BigQuery dataset.; Implement BigQuery Policy Tags with Dynamic Data Masking on the sensitive columns.

To meet the requirement of managing and rotating keys, Customer-Managed Encryption Keys (CMEK) via Cloud KMS is the correct approach. For column-level visibility control, BigQuery Policy Tags (Data Catalog) with Dynamic Data Masking is the standard solution. Authorized Views could work for row-level or table-level access but are less efficient for granular column masking compared to Policy Tags. Select TWO.

Policy tags allow for fine-grained access control and dynamic masking based on user roles.

Correct answer: Migrate to Dataproc using ephemeral clusters. Use a workflow template or Cloud Composer to spin up the cluster before the job starts and delete it immediately after completion.

For variable workloads that are idle for long periods, ephemeral Dataproc clusters are the most cost-effective solution. You only pay for the compute resources while the job is running. Persistent clusters would incur costs during the idle 20 hours. Migrating to Dataflow would require rewriting the Spark code, which is not specified as a desire. Compute Engine requires manual management and is less efficient.

Correct answer: Issue: Hotspotting if many sensors write simultaneously. Fix: Prefix the row key with a hash of the SensorID.

If SensorIDs are sequential or if specific sensors are much more active, writing sequentially can cause hotspotting on specific nodes. Hashing the SensorID distributes writes across the cluster. While `[SensorID]#[Timestamp]` is good for single-sensor queries, it doesn't solve the write distribution issue if IDs are close. However, the best answer for *general* hotspot prevention in this context is hashing or salting. Note: For the geographic query requirement, a different approach might be needed (like a secondary index or different table), but the primary issue with `[SensorID]` is often write distribution if not hashed.

Correct answer: Partition the table by `TransactionDate` and cluster by `StoreId`.

Partitioning by `TransactionDate` ensures that queries filtering on date only scan the relevant partitions, drastically reducing cost (bytes scanned). Clustering by `StoreId` sorts the data within each partition, which accelerates aggregations and filters on that column. This combination directly addresses both the filter and aggregation requirements.

Correct answer: BigQuery ML

BigQuery ML allows users to create and execute machine learning models in BigQuery using standard SQL queries. This fits the requirement of using SQL and avoids moving data out of the warehouse. Vertex AI Custom Training would require Python/container knowledge.

Correct answer: Enable Streaming Engine or increase the maximum number of workers.

High CPU utilization with increasing lag indicates the pipeline is under-provisioned for the volume of data or complexity of processing. Enabling Streaming Engine moves state storage and shuffle off the workers, reducing CPU load. Alternatively, increasing the max workers allows the autoscaler to provision more compute power. Switching to a larger machine type is also valid but requires a pipeline update/restart.