ADP Sample Questions & Answers

This statement is false. With CMEK, the customer manages the key within Google's Cloud Key Management Service (KMS). Google services (like Cloud Storage) still perform the cryptographic operations by requesting the KMS to use the key for encryption/decryption. The customer controls the key's lifecycle and access policies, but the key material resides within KMS. The scenario described (key held and used outside Google) applies to Customer-Supplied Encryption Keys (CSEK) or Cloud External Key Manager (EKM).

Cloud Data Fusion is the ideal service for this requirement. It is a fully managed, cloud-native data integration service that provides a graphical interface and a broad library of pre-built transformations, including those for data masking and de-identification. This aligns perfectly with the 'low-code' and 'graphical' requirements. Dataflow would require custom coding, and Cloud Composer is an orchestrator, not a data transformation tool itself.

For a regression model, mean_absolute_error (MAE) indicates the average absolute difference between the predicted values and the actual values. An MAE of 25000 means the predictions are off by an average of $25,000. The r2_score (R-squared) represents the proportion of the variance in the dependent variable (housing price) that is predictable from the independent variables. An r2_score of 0.85 means that the model's features explain 85% of the variability in housing prices, which is generally considered a good fit.

Cloud Composer is a managed Apache Airflow service. The primary interface for managing, monitoring, and troubleshooting DAGs (Directed Acyclic Graphs) is the Airflow web UI. This interface provides detailed views of DAG runs, task statuses, logs for individual tasks, trigger history, and the rendered code. While logs are also available in Cloud Logging and the environment can be managed in the Google Cloud Console, the Airflow UI is the specific tool designed for in-depth DAG troubleshooting.

The BigQuery Data Editor role is the most appropriate choice. It grants permissions to read, query, create, update, and delete tables within a dataset. This meets the requirements of running queries and creating new tables. It does not, however, grant permissions to delete the dataset itself or change its IAM policies, which are part of the BigQuery Data Owner role. BigQuery Data Viewer is too restrictive as it only allows reading data, not creating tables.

This configuration is the most cost-effective. After the initial 30 days of frequent access, the data becomes cold. Transitioning it directly to Archive storage, which is designed for long-term, infrequent access, will provide the lowest storage cost for the five-year retention period. A final action to delete the objects after 1825 days (365 * 5) fulfills the compliance requirement. Using Nearline or Coldline would be more expensive than Archive for this long-term storage use case.

This is the correct and standard way to execute a SQL query against BigQuery from a Python environment using the official client library. The SQL query correctly calculates the sum of purchase_amount for each customer_id, orders the results in descending order to find the top spenders, and limits the output to the top 5. The client.query(sql).to_dataframe() method executes the query and conveniently converts the results into a Pandas DataFrame for further analysis in the notebook.

This is the classic serverless pattern for lightweight, event-driven tasks on Google Cloud. A Cloud Function can be directly triggered by a Cloud Storage event (like a file upload). The function can execute the validation logic and then publish to Pub/Sub. This solution is fully serverless, scales to zero (making it cost-effective for infrequent events), and requires no infrastructure management. Dataflow is better suited for large-scale stream or batch processing, not single-file validation. Cloud Run is also a good serverless option but Cloud Functions are often simpler for direct event handling like this.

This query uses a window function ROW_NUMBER() to rank rows for each customer_id based on the last_modified_ts in descending order. This assigns a rank of 1 to the most recent record for each customer. The QUALIFY clause, specific to BigQuery and some other data warehouses, filters the results of the window function, keeping only the rows where the rank is 1. This is an efficient and standard pattern for deduplication in BigQuery. The other options use incorrect SQL syntax or logic for this task.