PDE Practice Test 27 — 15 Questions

Question 1

A team has trained a model using AutoML Tables. They want to deploy it for batch predictions on a schedule. What is the simplest approach?

Accepted Answer

Use Vertex AI Batch Prediction with a scheduled pipeline. Vertex AI Batch Prediction is the simplest approach because it is a managed service that directly supports batch predictions on AutoML Tables models without requiring additional infrastructure. By wrapping it in a scheduled Vertex AI pipeline, you can automate the entire workflow—triggering predictions on a schedule, handling input/output to Cloud Storage, and managing compute resources—all within the Vertex AI ecosystem, minimizing operational overhead.

Answer

Write a Cloud Function triggered by Cloud Scheduler

Answer

Export model to Cloud Storage and use Dataflow

Answer

Deploy to App Engine

Question 2

Which THREE Google Cloud services are typically used together in a production ML pipeline?

Accepted Answer

Cloud Storage. Cloud Storage is correct because it serves as the central artifact repository in a production ML pipeline on Google Cloud. It stores training data, model artifacts, and prediction inputs/outputs, enabling seamless integration with Vertex AI Training for model training and Vertex AI Prediction for serving. Without Cloud Storage, there is no durable, scalable, and cost-effective way to manage the large datasets and model binaries required for production ML workflows.

Answer

Cloud Functions

Answer

BigQuery

Question 3

A company runs a batch data processing workload using Dataproc clusters that are auto-scaled based on YARN memory utilization. During peak times, jobs take much longer than expected. Analysis shows the cluster is not scaling up despite high YARN memory utilization. What is the most likely cause?

Accepted Answer

The cluster is using primary workers only; auto-scaling only adds secondary workers. Dataproc clusters have two types of workers: primary workers (which run both HDFS and compute) and secondary workers (compute-only). The autoscaler can only add or remove secondary workers; it cannot scale primary workers. If the cluster uses only primary workers, the autoscaler has no secondary workers to add, so it cannot scale up even under high YARN memory utilization. This explains why the cluster remains static during peak times.

Answer

Spark dynamic allocation is disabled, preventing executors from using added workers

Answer

The cluster autoscaler is misconfigured to scale based on CPU, not memory

Answer

The autoscaler is set to scale down secondary workers, not up

Question 4

You manage a BigQuery reservation with 500 baseline slots and autoscaling up to 2000 slots. Your team runs a mix of interactive queries and batch load jobs. During peak hours, you notice that interactive queries are throttled when autoscaling slots are consumed by long-running batch loads. How can you ensure interactive queries get priority access to slots?

Accepted Answer

Create a separate reservation for interactive queries with a higher priority assignment.. BigQuery reservations allow you to create separate reservations for different workloads (e.g., interactive queries vs. batch loads) and assign them different priority levels. By creating a dedicated reservation for interactive queries with a higher priority, you ensure that interactive queries get preferential access to slots, even when autoscaling slots are consumed by long-running batch jobs. This directly addresses the contention issue without reducing overall capacity.

Answer

Reduce the baseline slots to 200 and rely solely on autoscaling.

Answer

Switch to on-demand pricing to eliminate slot contention.

Answer

Set the autoscaling max to 1000 slots for batch jobs.

Question 5

A team is training a large model using a custom container with TensorFlow on Vertex AI Training. They need to use multiple GPUs across several machines. Which strategy should they implement to maximize training throughput?

Accepted Answer

Use Vertex AI Training with a custom job specifying workerPoolSpecs and MultiWorkerMirroredStrategy. Vertex AI Training's custom job with workerPoolSpecs enables multi-machine, multi-GPU distributed training, and TensorFlow's MultiWorkerMirroredStrategy is specifically designed for synchronous distributed training across multiple workers. This combination maximizes throughput by efficiently synchronizing gradients across all GPUs on all machines using all-reduce communication, which is essential for large model training.

Answer

Use Cloud TPU Pods for distributed training

Answer

Use Dataflow for distributed training

Answer

Use a single worker with multiple GPUs and TensorFlow MirroredStrategy

Question 6

A startup is deploying a machine learning model for real-time fraud detection. They need low latency and automatic scaling during peak hours. Which Google Cloud service should they use?

Accepted Answer

Vertex AI Endpoints. Vertex AI Endpoints provide managed, autoscaling infrastructure designed for low-latency online predictions, making them ideal for real-time fraud detection. They automatically scale the number of compute nodes based on incoming traffic, ensuring peak-hour demand is met without manual intervention.

Answer

Cloud Functions

Answer

Batch Prediction on Vertex AI

Answer

Cloud AI Platform Prediction with custom containers

Question 7

A company has a production machine learning model deployed on Vertex AI Endpoint that predicts customer churn. The model is retrained weekly using a Vertex AI Pipeline that pulls new data from BigQuery. Recently, the model's accuracy has been declining. The data science team suspects data drift but is unsure. They have enabled Vertex AI Model Monitoring but have not set up any alerts. The team wants to diagnose and address the issue quickly. The pipeline runs successfully, and no errors are reported. The model endpoint is serving predictions with average latency of 200ms. What should the team do first?

Accepted Answer

Review Vertex AI Model Monitoring drift reports and set up alerts for significant drift. The team has already enabled Vertex AI Model Monitoring, which automatically tracks feature distributions and prediction statistics over time. The first diagnostic step should be to review the drift reports generated by Model Monitoring to confirm whether data drift is occurring, and then set up alerts so the team is proactively notified of significant drift in the future. This directly addresses the suspected root cause without unnecessary operational changes.

Answer

Immediately trigger a retraining pipeline with more recent data

Answer

Increase the number of replicas to reduce latency

Answer

Examine Cloud Logging for prediction errors

Question 8

A financial services company deploys a fraud detection model on Vertex AI using a custom prediction container that runs a PyTorch model. The model requires GPU acceleration. The deployment succeeds but predictions return an error: 'CUDA error: out of memory'. What should the team do to resolve this issue?

Accepted Answer

Increase the GPU machine type to one with more memory (e.g., from NVIDIA T4 to A100). The CUDA out-of-memory error indicates that the GPU's VRAM is insufficient to load the PyTorch model or process the inference batch. Increasing the GPU machine type to one with more memory, such as from an NVIDIA T4 (16 GB) to an A100 (40 or 80 GB), directly resolves the capacity issue. Vertex AI prediction endpoints allow you to select different accelerator types and sizes, and this change ensures the model fits within GPU memory.

Answer

Change the container to use a CPU-only image to avoid CUDA errors

Answer

Enable Vertex AI Model Monitoring to automatically scale the endpoint

Answer

Add CPU replicas to distribute the inferencing load

Question 9

A company runs Apache Spark jobs on Dataproc. They want to reduce costs by using preemptible instances for worker nodes. The jobs are fault-tolerant and can handle occasional node loss. However, the cluster must remain available for interactive querying during business hours. Which Dataproc cluster configuration meets these requirements?

Accepted Answer

Use a standard cluster with preemptible instances as secondary workers. Dataproc supports preemptible VMs as secondary workers, which are added to a cluster alongside non-preemptible primary workers. Secondary workers handle the bulk of shuffle and task execution, so if a preemptible node is reclaimed, the job can recover while the primary workers and master keep the cluster alive and available for interactive queries. This gives the cost benefit of preemptible pricing without risking cluster or master availability.

Answer

Use a single-node cluster that automatically scales with preemptible instances

Answer

Use standard cluster with master and worker nodes as preemptible instances

Answer

Use a high-availability cluster with preemptible instances for primary workers

Question 10

A company is building a real-time anomaly detection pipeline using Dataflow. Events are ingested from Pub/Sub, and the pipeline must compute a sliding window average every minute over a 1-hour window. Which TWO configurations are required for this pipeline? (Choose 2)

Accepted Answer

Set the pipeline to use event time for watermarking.. Option A is correct because a sliding window over event data must be based on event time so that late-arriving events are assigned to the correct window and watermarks track the progress of event time rather than the wall-clock time at which elements are processed. Option B is correct because the requirement is a 1-hour window that is recomputed every minute, which is exactly a SlidingWindow with a duration of 1 hour and a period (slide) of 1 minute. Option C is incorrect because a FixedWindow of 1 minute produces independent non-overlapping 1-minute aggregates, not a 1-hour average refreshed each minute. Option D is incorrect because stateful processing with a custom timer is a lower-level mechanism and is not required when Apache Beam's built-in sliding windows already express the desired semantics. Option E is incorrect because processing-time watermarking ignores the event timestamps and would misassign late or out-of-order Pub/Sub messages, breaking the anomaly detection logic.

Answer

Use a Fixed window of 1 minute.

Answer

Use stateful processing with a custom timer.

Answer

Set the pipeline to use processing time for watermarking.

Question 11

A data engineer needs to run an existing Spark job on Google Cloud with minimal code changes. The job requires Hive metastore access. Which Dataproc feature should they use to provide a managed Hive metastore?

Accepted Answer

Dataproc Metastore. Dataproc Metastore is a fully managed, highly available Hive metastore service (based on Hive Metastore 2.3/3.1) that can be attached to Dataproc clusters, providing a persistent, serverless metastore without running a separate Hive metastore on a cluster. It allows existing Spark jobs that require Hive metastore access to run with minimal code changes, since the metastore endpoint is configured via cluster properties.

Answer

Cloud SQL for MySQL

Answer

BigQuery as a Hive metastore

Answer

Dataproc on GKE

Question 12

Drag and drop the steps to create a Cloud Storage bucket with uniform bucket-level access into the correct order.

Accepted Answer

Navigate to Cloud Storage, click Create Bucket, enter bucket name, select Uniform under access control, click Create. Uniform bucket-level access simplifies permissions by using IAM policies at the bucket level instead of ACLs.

Answer

Click Create Bucket, navigate to Cloud Storage, enter bucket name, select Uniform, click Create

Answer

Navigate to Cloud Storage, click Create Bucket, select Uniform, enter bucket name, click Create

Answer

Navigate to Cloud Storage, click Create Bucket, enter bucket name, click Create, select Uniform

Question 13

A company is designing a data lake on Cloud Storage for analytics. They need to store data in various formats (Avro, Parquet, CSV) and enable efficient querying with BigQuery and Dataproc. Which THREE practices should they follow?

Accepted Answer

Use BigLake to create BigQuery tables that reference Cloud Storage data.. BigLake allows you to create BigQuery tables that reference data stored in Cloud Storage, enabling unified governance and fine-grained access control without moving data. This is essential for a data lake architecture where BigQuery and Dataproc need to query the same underlying data in various formats like Avro, Parquet, and CSV.

Answer

Disable encryption on the bucket to improve read performance.

Answer

Store all data in CSV format for simplicity.

Question 14

A company uses Cloud Storage as a data lake with raw, curated, and processed zones. Data in the raw zone should be automatically moved to a cheaper storage class after 30 days, and deleted after 1 year. What is the most efficient way to implement this?

Accepted Answer

Use Object Lifecycle Management with rules to transition to Coldline after 30 days and delete after 365 days.. Object Lifecycle Management in Cloud Storage allows you to set rules based on object age. You can transition objects to a lower-cost storage class (e.g., Nearline or Coldline) after 30 days and delete after 365 days.

Answer

Write a Cloud Function that runs daily, checks object ages, and moves/deletes them.

Answer

Use Cloud Scheduler to run a script that changes storage class and deletes objects.

Answer

Set a retention policy on the raw zone to prevent deletion and manually clean up.

Question 15

Drag and drop the steps to create a Cloud Composer environment for Apache Airflow into the correct order.

Accepted Answer

Enable Cloud Composer API → Create Cloud Storage bucket → Create Cloud Composer environment → Upload DAGs. Cloud Composer provides a managed Airflow environment for orchestrating workflows.

Answer

Create Cloud Storage bucket → Enable Cloud Composer API → Create Cloud Composer environment → Upload DAGs

Answer

Upload DAGs → Create Cloud Composer environment → Enable Cloud Composer API → Create Cloud Storage bucket

Answer

Create Cloud Composer environment → Enable Cloud Composer API → Create Cloud Storage bucket → Upload DAGs