Download Cloud Data Engineer Java for China India Philippines Complete Study and Certification Guid and more Exams Technology in PDF only on Docsity!
Philippines Complete Study and Certification
Guide
Question 1. Which Cloud Storage class is optimal for data accessed less than once a quarter but retained for several years? A) Standard B) Nearline C) Coldline D) Archive Answer: D Explanation: Archive offers the lowest storage cost for long-term, infrequently accessed data, with higher retrieval latency, fitting the described use case. Question 2. In a multi-regional BigQuery deployment, which benefit is most directly achieved? A) Lower network egress costs B) Automatic data replication across continents C) Sub-second query latency for all users D) Simplified IAM role management Answer: B Explanation: Multi-regional datasets are replicated across multiple regions, providing higher durability and locality for users in different geographies. Question 3. When choosing between Cloud SQL (MySQL) and Cloud Spanner for a banking transaction system requiring strong consistency and horizontal scaling, which is preferred? A) Cloud SQL (MySQL) B) Cloud Spanner C) Both are equally suitable D) Neither; use Bigtable
Philippines Complete Study and Certification
Guide
Answer: B Explanation: Cloud Spanner provides global strong consistency and horizontal scaling, essential for high-throughput, ACID-compliant banking workloads. Question 4. Which NoSQL option is best suited for a mobile gaming backend that stores user profiles as JSON documents with occasional complex queries? A) Bigtable B) Firestore in Native mode C) Cloud SQL D) Datastore in Datastore mode Answer: B Explanation: Firestore stores JSON-like documents, supports rich queries, and offers offline sync, making it ideal for mobile app back-ends. Question 5. In BigQuery, partitioning a table by DATE column primarily reduces cost by: A) Decreasing storage size B) Limiting the number of slots used per query C) Allowing queries to scan only relevant partitions D) Enabling automatic clustering Answer: C Explanation: Partition pruning lets queries scan only the partitions that satisfy the date filter, reducing the amount of data processed and thus cost. Question 6. Which schema design typically results in fewer joins and better query performance in a cloud-native data warehouse? A) Normalized (3NF) schema
Philippines Complete Study and Certification
Guide
Question 9. In Dataflow, what does the “Accumulation mode = DISCARDING_FIRED_PANES” mean for a trigger? A) All elements are retained after each pane fires B) Elements are removed after the pane fires, preventing re-emission C) The pipeline stops after the first pane fires D) Late data is ignored completely Answer: B Explanation: Discarding mode drops elements that have already been emitted, so subsequent panes only contain new data. Question 10. Which Cloud Pub/Sub feature enables exactly-once delivery semantics for Java subscribers? A) Ordering keys B) Acknowledgement deadline extension C) Exactly-once delivery (EOD) with dead-letter topics D) Message filtering Answer: C Explanation: Pub/Sub’s exactly-once delivery (EOD) guarantees that each message is processed only once, even if retries occur. Question 11. When migrating an on-prem Hadoop MapReduce job written in Java to Cloud Dataproc, which component can be reused without modification? A) HDFS file paths B) YARN resource manager settings C) MapReduce Java classes (Mapper, Reducer)
Philippines Complete Study and Certification
Guide
D) Hive metastore configurations Answer: C Explanation: The Java MapReduce classes are portable; they can run on Dataproc’s YARN cluster without code changes. Question 12. In Data Fusion, a custom Java plugin is needed to read a proprietary file format. Which SDK should you extend? A) SparkSource B) BatchSink C) Transform D) Connector Answer: C Explanation: The Transform plugin allows you to implement custom logic in Java to process input records, suitable for proprietary formats. Question 13. Which JDBC URL pattern is correct for connecting a Java application to a Cloud SQL PostgreSQL instance using the Cloud SQL Auth Proxy? A) jdbc:postgresql://127.0.0.1:5432/DB_NAME B) jdbc:postgresql:///DB_NAME? socketFactory=com.google.cloud.sql.postgres.SocketFactory C) jdbc:postgresql://::/DB_NAME D) jdbc:postgresql://cloudsql.googleapis.com/DB_NAME Answer: B Explanation: The Cloud SQL Auth Proxy uses a custom socket factory; the URL must specify the socketFactory class and the instance connection name.
Philippines Complete Study and Certification
Guide
Answer: A Explanation: The Cloud DLP API provides de-identification methods such as masking, tokenization, and redaction for PII. Question 17. In Cloud Composer, which of the following is the primary way to pass parameters from a DAG to a Java Dataflow job? A) XCom variables B) Environment variables in the worker VM C) Template fields in the DataflowPythonOperator D) JSON payload in the DataflowSubmitJobOperator Answer: D Explanation: The DataflowSubmitJobOperator accepts a JSON payload that defines job parameters, which are then passed to the Java job. Question 18. Which Terraform resource creates a BigQuery dataset with a default table expiration of 90 days? A) google_bigquery_table B) google_bigquery_dataset C) google_bigquery_job D) google_bigquery_dataset_iam_policy Answer: B Explanation: The google_bigquery_dataset resource includes a default_table_expiration_ms attribute to set expiration for tables. Question 19. During a Dataflow job, the “System Lag” metric indicates: A) Time spent waiting for container startup
Philippines Complete Study and Certification
Guide
B) Difference between processing time and event time C) Network latency between workers D) Time taken for JVM garbage collection Answer: B Explanation: System Lag measures how far behind the pipeline is from the event timestamps, reflecting processing delay. Question 20. Which Cloud Monitoring alerting policy condition would best detect a sudden drop in Dataflow worker CPU utilization? A) Metric absence for “worker/total_cpu_time” B) Threshold condition on “worker/cpu_utilization” < 10% for 5 minutes C) Rate of change on “worker/total_bytes_processed” > 0 D) Log-based metric for “worker/exception” occurrences Answer: B Explanation: A threshold condition on low CPU utilization over a short period flags under-utilized workers. Question 21. When preparing a dataset for Vertex AI AutoML, which data format is recommended for tabular data? A) CSV with header row B) JSON Lines without schema C) Avro with embedded schema D) Parquet with partitioning Answer: A Explanation: AutoML Tabular expects CSV files with a header row that defines column names and types.
Philippines Complete Study and Certification
Guide
B) com.google.cloud.vision.v1.LabelDetectionClient C) com.google.cloud.vision.ImageClient D) com.google.cloud.vision.v1.AnnotateImageRequest Answer: A Explanation: ImageAnnotatorClient provides methods such as batchAnnotateImages for label detection. Question 25. For a cost-optimized Dataflow pipeline in India, which worker type should you select? A) n1-standard-4 preemptible B) n1-highmem- C) n2-standard- D) n1-standard-1 non-preemptible Answer: A Explanation: Preemptible VMs are up to 80 % cheaper; they are suitable for fault-tolerant pipelines and reduce cost. Question 26. In a Cloud Spanner database, which schema design avoids hotspotting for time-series data? A) Using a monotonically increasing integer as the primary key B) Prefixing the timestamp with a hash suffix C) Storing timestamps in reverse chronological order D) Using only the timestamp as the primary key Answer: B Explanation: Adding a hash or random prefix distributes writes across multiple splits, preventing hotspotting.
Philippines Complete Study and Certification
Guide
Question 27. Which Bigtable row key pattern best supports efficient range scans for IoT sensor data? A) sensorId_timestamp B) timestamp_sensorId C) sensorId#timestamp (reverse) D) timestamp#sensorId (reverse) Answer: A Explanation: Placing sensorId first groups rows per device, while the timestamp allows range scans within each sensor’s data. Question 28. When configuring Cloud Storage lifecycle rules for a dataset that becomes cold after 30 days, which action should be set? A) Delete after 30 days B) Transition to Nearline after 30 days C) Transition to Coldline after 30 days D) Set retention policy of 30 days Answer: C Explanation: Coldline is intended for data accessed less than once a quarter, matching the “cold after 30 days” scenario. Question 29. Which of the following is a primary advantage of using Cloud Data Fusion over hand-coded Beam pipelines? A) Lower runtime latency B) Visual UI for building pipelines with drag-and-drop C) Automatic generation of Java code for all transforms
Philippines Complete Study and Certification
Guide
Question 32. When using Terraform to provision a Cloud Composer environment, which argument defines the number of Airflow workers? A) node_count B) scheduler_count C) worker_count D) airflow_worker_number Answer: C Explanation: The worker_count attribute sets the size of the Airflow worker pool in the Composer environment. Question 33. Which feature of Cloud KMS enables automatic rotation of CMEK for BigQuery tables every 90 days? A) Key versioning with a rotation schedule policy B) Manual key version creation via API C) Auto-generation of new keys per dataset D) Integration with Cloud Scheduler only Answer: A Explanation: KMS allows you to define a rotation period; new key versions are created automatically, and services like BigQuery can be configured to use the latest version. Question 34. **In Vertex AI Feature Store, what is the purpose of “feature lineage”? ** A) Tracking which model used a feature B) Recording the source of feature values and transformations C) Versioning the schema of a feature group
Philippines Complete Study and Certification
Guide
D) Monitoring feature usage frequency Answer: B Explanation: Feature lineage provides traceability from raw data through transformations to the stored feature values. Question 35. Which Java exception is thrown by the Cloud SQL Auth Proxy when the connection to the instance cannot be established? A) IOException B) SQLTransientConnectionException C) CloudSqlConnectionException D) IllegalStateException Answer: B Explanation: The proxy surfaces transient connection failures as SQLTransientConnectionException, indicating a temporary inability to connect. Question 36. When designing a BigQuery table for clickstream logs, which partitioning method minimizes cost while supporting queries by event date? A) Ingestion-time partitioning B) Integer range partitioning on session_id C) Date column partitioning on event_date D) No partitioning, rely on clustering only Answer: C Explanation: Partitioning by the event_date column allows queries to prune partitions based on date filters, reducing processed data. Question 37. Which Cloud Dataflow runner is required to execute a pipeline that uses Apache Beam’s Python SDK from a Java project?
Philippines Complete Study and Certification
Guide
Explanation: request.auth.claims.email contains the user's email address, allowing condition-based access control for a domain. Question 40. When using Cloud DLP to de-identify credit card numbers, which transformation type should you select? A) CryptoDeterministicConfig B) RedactConfig C) ReplaceWithInfoTypeConfig D) FixedSizeBucketingConfig Answer: B Explanation: RedactConfig removes the matched sensitive data entirely from the output. Question 41. Which Dataflow metric indicates the number of elements that have been dropped due to late-data handling? A) element_count B) dropped_elements_count C) late_data_count D) watermarks_lag Answer: B Explanation: The dropped_elements_count metric tracks elements discarded because they arrived after the allowed lateness. Question 42. In a Java Beam pipeline, which transform is most appropriate for deduplicating records based on a unique key? A) GroupByKey B) Distinct
Philippines Complete Study and Certification
Guide
C) Combine.perKey D) Filter.byKey Answer: B Explanation: Distinct removes duplicate elements; when combined with a key extraction function, it deduplicates by that key. Question 43. Which of the following best describes the purpose of “Flex Slots” in BigQuery? A) Permanent on-demand slots that never expire B) Short-term, pre-emptible slots that can be purchased for up to 24 hours C) Reserved slots that guarantee capacity for a year D) Slots used exclusively for streaming inserts Answer: B Explanation: Flex Slots are short-lived, on-demand slots that can be added for a limited period, offering cost flexibility. Question 44. When using Vertex AI to train a custom image classification model, which data format must the training images be stored in? A) CSV with image URLs B) TFRecord files with encoded images C) Cloud Storage folder with labeled sub-folders D) Parquet with image binary column Answer: C Explanation: Vertex AI AutoML image classification expects images organized in Cloud Storage with one sub-folder per label.
Philippines Complete Study and Certification
Guide
Answer: C Explanation: STRUCT and ARRAY allow storage of nested records, and Standard SQL can query them directly. Question 48. Which Dataflow windowing strategy is most suitable for calculating a rolling 7-day sum that updates every hour? A) Fixed window of 7 days B) Sliding window of 7 days with 1-hour slide C) Session window with 7-day gap D) Global window with trigger every hour Answer: B Explanation: A sliding window of 7 days that slides every hour produces the desired rolling aggregation. Question 49. Which IAM role is required for a service account to create and delete Cloud Composer environments? A) roles/composer.admin B) roles/composer.environmentEditor C) roles/composer.user D) roles/editor Answer: A Explanation: The composer.admin role includes permissions to manage Composer environments. Question 50. In a Java Dataflow job, which method should you override to perform cleanup after all elements have been processed? A) startBundle()
Philippines Complete Study and Certification
Guide
B) processElement() C) finishBundle() D) teardown() Answer: C Explanation: finishBundle() is called after a bundle of elements is processed, allowing finalization logic. Question 51. When using BigQuery’s “CREATE TABLE … CLONE” statement, which of the following is true? A) The clone shares the same underlying storage as the source table. B) The clone is a deep copy that consumes additional storage. C) Cloned tables cannot be partitioned. D) Cloning is only allowed for external tables. Answer: A Explanation: Cloned tables are metadata copies that reference the same data blocks, incurring no extra storage until modified. Question 52. Which of the following best describes “hotspotting” in Cloud Bigtable? A) Excessive read latency due to large row sizes B) Uneven distribution of writes to a small set of tablets C) Over-provisioned nodes causing idle CPU D) Frequent schema changes causing compaction delays Answer: B Explanation: Hotspotting occurs when many writes target the same tablet range, leading to performance bottlenecks.
