PDE Practice Test 46 — 15 Questions

Question 1

Which Google Cloud service provides a serverless Spark environment where you can run Spark jobs without provisioning or managing a cluster?

Accepted Answer

Dataproc Serverless. Dataproc Serverless allows you to submit Spark jobs that run on auto-scaled infrastructure without cluster management. Dataflow is for Beam pipelines. Dataprep is for data wrangling. Data Fusion is for visual ETL.

Answer

Dataflow

Answer

Dataprep

Answer

Cloud Data Fusion

Question 2

A data engineer needs to load 10 TB of CSV files from Amazon S3 into Google BigQuery on a daily basis. Which service should they use to automate this transfer?

Accepted Answer

BigQuery Data Transfer Service. Storage Transfer Service can transfer data from Amazon S3 to Google Cloud Storage, but it does not load directly into BigQuery. BigQuery Data Transfer Service can import from Amazon S3 directly into BigQuery tables. Other options are not suitable: Cloud Data Fusion is for ETL pipelines, not simple transfer; Transfer Appliance is for offline petabyte-scale transfers; Dataproc is for Spark/Hadoop jobs.

Answer

Dataproc

Answer

Cloud Data Fusion

Answer

Storage Transfer Service

Question 3

You are migrating a large on-premises data warehouse to BigQuery. The data includes sensitive PII columns that must be masked for certain users. Which BigQuery feature can automatically redact PII in query results based on user roles?

Accepted Answer

Column-level security with data masking. BigQuery column-level security with policy tags and data masking can automatically mask sensitive data based on IAM roles. Authorized views require manual creation. Dynamic data masking is part of column-level security. IAM conditions don't mask data.

Answer

IAM conditions on tables

Answer

Authorized views

Answer

Cloud DLP API

Question 4

A healthcare company streams patient monitoring data to Cloud Pub/Sub. A Dataflow pipeline reads the stream, enriches with patient records from BigQuery, and writes to Bigtable for real-time queries. The BigQuery lookup is slow and causes pipeline lag. What is the best approach to improve performance?

Accepted Answer

Use a side input from a BigQuery query with a global window and periodic refresh.. Using a side input from BigQuery with a global window and periodic refresh allows the Dataflow pipeline to cache the patient records in memory across all workers, avoiding per-element slow lookups. This pattern leverages Beam's side input semantics to broadcast a relatively static lookup table, significantly reducing latency compared to synchronous BigQuery queries for each incoming event.

Answer

Increase the number of Dataflow workers and use vertical scaling.

Answer

Use BigQuery's streaming read API in the pipeline.

Answer

Pre-join the data in a batch pipeline and load into Bigtable.

Question 5

A data engineer needs to design a batch processing pipeline using Cloud Data Fusion. The pipeline should read data from Cloud Storage, perform transformations (join, filter, aggregate), and write to BigQuery. What is the most efficient way to handle the transformations?

Accepted Answer

Use Data Fusion Wrangler to visually design the transformations and then run the pipeline on a Dataproc cluster.. Cloud Data Fusion Wrangler provides a visual, no-code interface for designing transformations (join, filter, aggregate) that are then compiled into an Apache Spark or MapReduce program and executed on a Dataproc cluster. This approach leverages Data Fusion's native integration with Dataproc for efficient, scalable batch processing without manual coding, while keeping the pipeline fully managed within the Data Fusion ecosystem.

Answer

Use SQL queries in BigQuery to perform the transformations after loading raw data into staging tables.

Answer

Use custom Python scripts in a Cloud Function triggered after the files land in Cloud Storage.

Answer

Use Apache Spark on Dataproc to code the transformations manually, bypassing Data Fusion.

Question 6

A company needs to store transactional data for a global customer base with strong consistency and 99.999% availability SLA. They anticipate millions of transactions per day across multiple regions. Which TWO storage options meet these requirements? (Choose 2)

Accepted Answer

Firestore (in Datastore mode). Cloud Spanner and Firestore (in Datastore mode) both provide globally distributed, strongly consistent transactions with a 99.999% SLA. Cloud Spanner offers a relational model with ACID across regions, while Firestore (Datastore mode) provides a schemaless document database with strong consistency and multi-region replication. Cloud Bigtable does not offer strong consistency across rows. AlloyDB and Cloud SQL with HA are regional services, not global, and do not meet the 99.999% SLA across multiple regions.

Answer

Cloud Bigtable

Answer

AlloyDB

Answer

Cloud SQL (with HA)

Question 7

A data engineer needs to store transactional data for an e-commerce application that requires ACID compliance, automatic failover, and point-in-time recovery. The expected throughput is a few thousand transactions per second. Which Google Cloud storage option should they choose?

Accepted Answer

Cloud SQL. Cloud SQL is the correct choice because it provides full ACID compliance, automated failover with high availability configurations, and point-in-time recovery via binary log replay. It supports up to several thousand transactions per second with appropriate machine sizing, making it suitable for this e-commerce workload.

Answer

Firestore

Answer

Cloud Bigtable

Answer

Cloud Spanner

Question 8

A Dataflow streaming pipeline uses stateful transformations with per-key state and timers. After a deployment, the team observes that the pipeline is reprocessing events from the last 30 minutes every time it restarts. The pipeline's checkpoint is configured to persist every 10 seconds. Which change should be made to prevent unnecessary reprocessing?

Accepted Answer

Enable idempotent writes to the sink by adding a unique identifier per event.. Enabling idempotent writes ensures that even if events are reprocessed due to pipeline restarts, the sink will deduplicate them based on the unique identifier. This prevents duplicate data from being written, which is the core issue when stateful transformations cause reprocessing of events from the last 30 minutes. The checkpoint interval (10 seconds) is already frequent enough; the problem is not checkpoint frequency but the lack of deduplication at the sink.

Answer

Use a non-volatile state backend like Cloud Bigtable for state storage.

Answer

Increase the checkpoint interval to 60 seconds to reduce frequency of checkpoints.

Answer

Decrease the checkpoint interval to 1 second to checkpoint more frequently.

Question 9

Which TWO actions should you take to ensure model reliability in a production Vertex AI Endpoint?

Accepted Answer

Monitor prediction accuracy in production with logging and alerts. Monitoring prediction accuracy with logging and alerts (B) is essential for detecting model drift, data drift, and performance degradation in production. Vertex AI provides model monitoring features that automatically log prediction requests and responses, compute statistics, and trigger alerts when skew or drift thresholds are breached, enabling proactive remediation.

Answer

Use only batch predictions to avoid real-time issues

Answer

Disable request/response logging to reduce latency

Answer

Use a single model endpoint for all traffic

Question 10

A company uses Pub/Sub to decouple services. They have a topic with two subscriptions: Subscription A is a push subscription that sends messages to a Cloud Function; Subscription B is a pull subscription used by a Dataflow job. They need to ensure that messages are processed in order for a specific device_id. Which TWO configurations should they apply?

Accepted Answer

Enable message ordering on the topic and set an ordering key for each message.. Enabling message ordering on the topic and setting an ordering key (e.g., device_id) ensures that messages with the same key are delivered to subscribers in the order they were published. This is a fundamental Pub/Sub feature that guarantees FIFO (first-in, first-out) delivery per ordering key, which directly addresses the requirement for processing messages in order for a specific device_id.

Answer

Disable duplicate filtering on the topic.

Answer

Configure the Cloud Function to retry on failure with exponential backoff.

Answer

Use a single subscription with multiple concurrent consumers.

Question 11

A company runs a production Dataflow streaming pipeline that reads from Pub/Sub, groups events by customer ID, and writes to BigQuery. The pipeline uses global windows with triggers. After a recent code change, the pipeline started generating duplicate events in BigQuery for the same customer ID. The previous version did not have duplicates. The team reviews the code and sees that the trigger was changed from 'afterProcessingTime' to 'afterWatermark'. What is the most likely reason for duplicates?

Accepted Answer

Late-arriving events cause the afterWatermark trigger to fire additional panes for the same window. The change from `afterProcessingTime` to `afterWatermark` introduces a dependency on the watermark, which estimates event time progress. When late-arriving events (those with timestamps before the watermark) arrive after the watermark has advanced, the `afterWatermark` trigger fires an additional pane for the same window, causing duplicate writes to BigQuery. The previous trigger (`afterProcessingTime`) fired based on processing time, which does not react to late data in the same way, hence no duplicates.

Answer

The afterProcessingTime trigger fired multiple times for the same window

Answer

The pipeline is firing early and on-time panes for the same window

Answer

The pipeline uses accumulation mode which accumulates results across firings

Question 12

A retail company uses Vertex AI Pipelines to automate monthly retraining of a recommendation model. The pipeline consists of three steps: (1) extract data from BigQuery, (2) train a TensorFlow model on Vertex AI Training, (3) upload the model to Vertex AI Model Registry and deploy to an endpoint if performance metrics improve. Recently, the pipeline has been failing at step 2 with the error: 'The job was cancelled by the system because it exceeded the maximum training time of 3600 seconds.' You have confirmed that the training code is correct and the data size has not changed significantly. What should you do to fix this pipeline failure? A) Reconfigure the pipeline to use a larger machine type for training. B) Set the training timeout to 7200 seconds in the pipeline configuration. C) Reduce the training dataset size by sampling fewer rows. D) Switch from TensorFlow to a simpler model framework.

Accepted Answer

Set the training timeout to 7200 seconds in the pipeline configuration.. The default timeout for a training job in Vertex AI Pipelines is 3600 seconds; increasing the timeout allows the job to complete. Option A (larger machine) may help but is not a direct fix for timeout. Option C (reducing data) degrades model quality. Option D (changing framework) is drastic and unnecessary.

Answer

Reduce the training dataset size by sampling fewer rows.

Answer

Switch from TensorFlow to a simpler model framework.

Answer

Reconfigure the pipeline to use a larger machine type for training.

Question 13

A company uses Cloud Pub/Sub for a real-time data pipeline. The subscription has a backlog of millions of messages that are not being processed quickly enough. In Cloud Monitoring, you observe that the 'subscription/num_undelivered_messages' metric is high and growing, while 'subscription/oldest_unacked_message_age' is also increasing. Which action is MOST likely to reduce the backlog?

Accepted Answer

Increase the number of subscribers or the throughput capacity of the existing subscribers.. The backlog indicates that subscribers cannot keep up with the message flow. Increasing the number of subscribers or scaling their throughput capacity directly addresses the processing bottleneck, allowing messages to be pulled and acknowledged faster. Cloud Pub/Sub scales horizontally, so adding more pull subscribers or increasing the resources of existing ones (e.g., more worker threads, higher CPU/memory) reduces the backlog.

Answer

Delete the subscription and recreate it with a larger message retention duration.

Answer

Reduce the acknowledgment deadline to force faster processing.

Answer

Change the subscription type from push to pull.

Question 14

Drag and drop the steps to migrate an on-premises MySQL database to Cloud SQL using Database Migration Service into the correct order.

Accepted Answer

First, prepare the source database by enabling binary logging and creating a replication user. Second, create a target Cloud SQL instance. Third, create a connection profile in Database Migration Service. Fourth, create and start the migration job, then perform the final cutover (promote Cloud SQL).. Database Migration Service enables minimal-downtime migrations using replication.

Answer

First, create a target Cloud SQL instance. Second, prepare the source database by enabling binary logging and creating a replication user. Third, create a connection profile in Database Migration Service. Fourth, create and start the migration job, then perform the final cutover.

Answer

First, prepare the source database by enabling binary logging and creating a replication user. Second, create a connection profile in Database Migration Service. Third, create a target Cloud SQL instance. Fourth, create and start the migration job, then perform the final cutover.

Answer

First, prepare the source database by enabling binary logging and creating a replication user. Second, create and start the migration job, then perform the final cutover. Third, create a target Cloud SQL instance. Fourth, create a connection profile in Database Migration Service.

Question 15

You configured a model deployment monitor on your Vertex AI endpoint as shown. What will happen when the feature 'age' has a skew of 0.4?

Accepted Answer

An alert will be sent to admin@example.com. The monitoring configuration shows an alert threshold of 0.3 for the feature 'age', and a skew of 0.4 exceeds that threshold. Vertex AI Model Monitoring will trigger the configured alert action, which in this case is sending an email to admin@example.com. The alert is based on the specific feature's threshold, not on any other feature's threshold.

Answer

The endpoint will automatically roll back to a previous model version

Answer

No alert will be sent because the skew threshold is 0.2 for income

Answer

An alert will be sent only if both features exceed their thresholds