PDE Practice Exam 55 — 20 Questions

Question 1

A startup is deploying a PyTorch model on Google Cloud. They need to serve predictions for a mobile app with bursty traffic. Which service is most cost-effective?

Accepted Answer

Vertex AI Endpoints with autoscaling and a minimum of 0 replicas to scale down to zero. Most cost-effective because Vertex AI Endpoints with autoscaling and minimum of 0 replicas can scale down to zero when idle, eliminating cost during no traffic. This is ideal for bursty traffic patterns. Option B has minimum replicas, incurring cost even when idle. Option C (App Engine manual scaling) does not scale to zero easily. Option D (Cloud Run with CPU always allocated) charges for CPU even when idle, making it less cost-effective for bursty traffic.

Answer

Vertex AI Endpoints with a minimum number of replicas

Answer

App Engine with manual scaling

Answer

Cloud Run with CPU always allocated

Question 2

Refer to the exhibit. A subscriber is unable to pull messages from the topic. What is the most likely cause?

Accepted Answer

No subscription has been created for the topic.. In Google Cloud Pub/Sub, a topic is a named resource to which messages are sent by publishers. Subscribers must create a subscription (pull or push) to receive messages from that topic. If no subscription exists, the subscriber cannot pull any messages because there is no delivery endpoint or pull queue attached to the topic. Option C correctly identifies this missing subscription as the root cause.

Answer

The service account has the subscriber role but the topic is not configured correctly.

Answer

The service account needs roles/pubsub.viewer to list subscriptions.

Answer

The service account lacks roles/pubsub.publisher.

Question 3

A data pipeline processes streaming data with Dataflow. The team notices occasional data duplication in BigQuery. What is the best approach to ensure exactly-once processing?

Accepted Answer

Configure Dataflow with exactly-once sinks using file staging and deduplication.. Dataflow's exactly-once sinks use a two-phase commit protocol with file staging and deduplication to ensure that each record is written exactly once to the sink, even if the pipeline retries. This approach handles the inherent at-least-once delivery from Pub/Sub by staging output files and committing them atomically, preventing duplicates in BigQuery without relying on downstream deduplication.

Answer

Use Pub/Sub with at-least-once delivery and deduplicate in BigQuery using a unique identifier.

Answer

Use Cloud Functions to deduplicate messages before they enter the pipeline.

Answer

Enable idempotent writes in BigQuery.

Question 4

A team deploys a new version of a Cloud Function. After deployment, error rates increase significantly. What is the most efficient way to diagnose the cause?

Accepted Answer

Check Cloud Logging for error stacks and exceptions.. Cloud Logging automatically captures error stacks and exceptions from Cloud Functions without requiring code changes. Checking these logs is the most efficient first step because it provides immediate visibility into the root cause of errors, such as unhandled exceptions, timeouts, or dependency failures, without incurring additional deployment overhead.

Answer

Deploy a debug version with additional logging.

Answer

Increase the function timeout and retry settings.

Answer

Immediately rollback to the previous version.

Question 5

A company uses Cloud Monitoring to track application latency. They notice a spike in latency every 30 minutes. What is the best initial step to diagnose the issue?

Accepted Answer

Check if scheduled jobs or cron tasks overlap.. Recurring spikes at regular intervals often indicate a scheduled process (e.g., cron job, batch job) that runs every 30 minutes. Checking for overlapping scheduled jobs is the most efficient first step before scaling or other actions.

Answer

Increase the number of instances to handle the load.

Answer

Enable Cloud Trace for all requests.

Answer

Change the alert threshold to ignore the spikes.

Question 6

You manage a team that deploys multiple versions of a computer vision model for A/B testing on Vertex AI Endpoints. You need to route a small percentage of traffic to a canary version while the rest goes to the stable version. You also need to gradually increase the canary traffic over time based on performance metrics. Which approach should you take?

Accepted Answer

Deploy both models to the same endpoint and configure traffic splitting percentages using the Vertex AI console or API.. Vertex AI Endpoints natively support traffic splitting between model versions deployed to the same endpoint. This allows you to assign a percentage of traffic (e.g., 5%) to a canary version and the remainder to the stable version, and then adjust the split over time via the console or API as performance metrics dictate. This approach avoids the complexity and latency of external load balancers or application-level routing.

Answer

Create two separate endpoints, one for each version, and use a separate load balancer to route a percentage of requests to the canary endpoint.

Answer

Use Cloud Armor with weighted backend services to route a portion of requests to the canary version.

Answer

Implement feature flags in the application code to randomly select the model version for each prediction request.

Question 7

The exhibit shows an IAM policy for a BigQuery dataset. A Dataflow job is failing with 'Access Denied: Table ... User does not have bigquery.tables.get permission'. Which additional role should be granted to the service account?

Accepted Answer

roles/bigquery.dataEditor. The error indicates the service account lacks the `bigquery.tables.get` permission, which is required to read table metadata. `roles/bigquery.dataEditor` includes this permission along with `bigquery.tables.get`, `bigquery.tables.update`, and `bigquery.tables.export`, making it the minimal role that resolves the access denied error for a Dataflow job reading from a BigQuery table.

Answer

roles/bigquery.admin

Answer

roles/bigquery.user

Answer

roles/bigquery.jobUser

Question 8

A company has a Dataflow pipeline that reads from Pub/Sub, applies transformations, and writes to BigQuery. The pipeline is failing with 'deadline exceeded' errors during peak hours. The team suspects that the pipeline cannot keep up with the incoming data rate. They also notice that the autoscaling algorithm sets maxNumWorkers to 10, but the pipeline only scales to 5 workers. What is the most likely cause of the inadequate scaling?

Accepted Answer

The pipeline is CPU-bound and the autoscaler evaluates that adding more workers would not improve throughput. The autoscaler in Dataflow evaluates CPU utilization and throughput per worker. If the pipeline is CPU-bound, adding more workers does not reduce per-worker CPU load or improve throughput, so the autoscaler stops at 5 workers even though maxNumWorkers is 10. This is a classic symptom of a bottleneck that cannot be parallelized further, such as a single-threaded transformation or a hot key in a GroupByKey operation.

Answer

The maxNumWorkers setting is too low and should be reduced to trigger more aggressive scaling

Answer

BigQuery streaming quota is limiting the number of concurrent writes

Answer

The Pub/Sub subscription has a per-subscriber throughput limit of 5 workers

Question 9

A company uses Cloud Dataproc to run Spark jobs on ephemeral clusters. The input data is in Cloud Storage and output is also to Cloud Storage. The cluster is created and deleted daily. The cost is high due to spinning up nodes. Which change can reduce cost without sacrificing performance?

Accepted Answer

Use a combination of standard and preemptible VMs for worker nodes. Using a combination of standard and preemptible VMs for worker nodes reduces cost significantly while maintaining performance. Preemptible VMs are up to 80% cheaper than standard VMs, and since Spark is fault-tolerant and can handle node preemptions via speculative execution, the job can complete without performance degradation. Standard VMs for master nodes ensure cluster stability, while preemptible workers handle the bulk of data processing.

Answer

Use standard VMs with a larger number of smaller machines

Answer

Use Cloud Dataflow instead

Answer

Use preemptible VMs for all nodes

Question 10

A healthcare organization is deploying a model that processes protected health information (PHI). They need to ensure that the inference data is encrypted in transit and at rest, and access is audited. Which combination of services meets these requirements?

Accepted Answer

Vertex AI Endpoints with VPC-SC, CMEK, and Cloud Audit Logs. VPC Service Controls (VPC-SC) provides data exfiltration protection and ensures inference data remains within a defined security perimeter, Customer-Managed Encryption Keys (CMEK) encrypt data at rest with keys controlled by the organization, and Cloud Audit Logs capture all access events for auditing. This combination directly addresses encryption in transit (via VPC-SC perimeter enforcement) and at rest (via CMEK), plus access auditing via Cloud Audit Logs.

Answer

Cloud Run with VPC connector and Cloud KMS

Answer

Vertex AI Endpoints with IAM and Cloud Monitoring

Answer

AI Platform Prediction with Cloud Armor

Question 11

A healthcare company uses Vertex AI to deploy a medical image classification model. The model is deployed on a private endpoint with automatic scaling (minReplicaCount=2, maxReplicaCount=10). The model uses a custom container with a GPU for inference. Recently, during peak business hours (9 AM - 5 PM), users report that prediction requests frequently time out after 60 seconds, and the error rate increases. The team checks Cloud Monitoring and observes that CPU utilization averages 40%, GPU utilization averages 30%, and the number of replicas stays at 2. There are no errors in the container logs. The model serves a few hundred requests per second during peak. The team suspects the issue is not resource saturation but something else. What should they do to resolve the problem?

Accepted Answer

Optimize the prediction container to handle requests faster by reducing image pre-processing and using async I/O.. The symptoms—low CPU/GPU utilization, replicas stuck at 2, and timeouts—indicate that the container is taking too long to process each request, not that resources are saturated. Optimizing the container (e.g., reducing image pre-processing, using async I/O) reduces per-request latency, allowing the model to handle the same request rate within the 60-second timeout. This directly addresses the root cause without changing scaling or timeout settings.

Answer

Switch from online prediction to batch prediction using Vertex AI Batch Prediction.

Answer

Increase the minReplicaCount to 5 to ensure more replicas are always available.

Answer

Increase the request timeout setting on the load balancer to 120 seconds.

Question 12

You are optimizing a Dataflow pipeline that performs a group-by-key transformation on a large, skewed dataset. The pipeline is experiencing high latency due to data skew (some keys have many more values). Which TWO actions can help mitigate the skew? (Choose two.)

Accepted Answer

Use hot key detection and split the hot key into multiple sub-keys (e.g., append a random number).. Splitting a hot key into multiple sub-keys (e.g., by appending a random number) distributes the values across multiple shards during the shuffle phase, reducing the load on any single worker. This technique, often called "salting," is a standard pattern in Dataflow and Apache Beam to handle data skew by breaking the bottleneck caused by a single key with disproportionately many values.

Answer

Enable the Dataflow service's automatic reshuffling feature.

Answer

Use CoGroupByKey to reduce the number of keys.

Answer

Increase the number of worker machines.

Question 13

Your team uses Cloud Dataproc to run a Spark ML training job. The job is failing with an error: 'Container killed by YARN for exceeding memory limits.' What should you do to fix this?

Accepted Answer

Increase the spark.executor.memory property. The error 'Container killed by YARN for exceeding memory limits' indicates that the Spark executor process is using more memory than the YARN container allows. Increasing `spark.executor.memory` allocates a larger YARN container for each executor, providing the necessary headroom for the Spark application's memory demands, including overhead for off-heap memory and JVM internals.

Answer

Use preemptible VMs for faster execution

Answer

Increase the number of worker nodes

Answer

Enable the external shuffle service

Question 14

Your MLOps pipeline uses Vertex AI Pipelines. You want to ensure that model training uses a consistent environment with specific Python package versions. Which approach best achieves this?

Accepted Answer

Build a custom container image with all dependencies and use it in the training step. Building a custom container image with all dependencies ensures a fully deterministic and reproducible environment for model training. Vertex AI Pipelines executes each step as a container, so by pre-installing specific Python package versions into a custom image, you eliminate any risk of version drift or network issues during package installation at runtime. This approach aligns with MLOps best practices for environment consistency and is the most reliable method when exact package versions are critical.

Answer

Include a requirements.txt file in the pipeline step and let Vertex AI install them

Answer

Use a pre-built deep learning container from Deep Learning Containers and install packages at runtime

Answer

Specify the Python version and package versions in the training job configuration

Question 15

Drag and drop the steps to configure a VPC network with private Google access for on-premises connectivity using Cloud VPN into the correct order.

Accepted Answer

Create VPC network with subnets, then enable Private Google Access on the subnets, then create Cloud VPN gateway and VPN tunnel, then configure the on-premises VPN device.. Private Google Access allows on-premises hosts to reach Google APIs via VPN without public IPs.

Answer

Create VPC network with subnets, then create Cloud VPN gateway and VPN tunnel, then enable Private Google Access on the subnets, then configure the on-premises VPN device.

Answer

Enable Private Google Access on the subnets, then create VPC network with subnets, then create Cloud VPN gateway and VPN tunnel, then configure the on-premises VPN device.

Answer

Create Cloud VPN gateway and VPN tunnel, then create VPC network with subnets, then enable Private Google Access on the subnets, then configure the on-premises VPN device.

Question 16

Your team needs to store time-series data from millions of IoT devices. Each device sends a reading every 5 minutes, and the total data volume is about 2 TB per month. The most common query pattern is retrieving all readings for a specific device over a time range (e.g., last 24 hours). Which storage service should you choose?

Accepted Answer

Cloud Bigtable. Cloud Bigtable is a fully managed, scalable NoSQL database designed for high-throughput, low-latency time-series data. It supports single-row key lookups and range scans, making it ideal for retrieving all readings for a specific device over a time range (e.g., last 24 hours) from millions of IoT devices generating 2 TB/month. Its row key design (e.g., device_id + timestamp) enables efficient time-range queries without full table scans, unlike object storage or analytical warehouses.

Answer

Cloud Storage (objects per device per time interval)

Answer

BigQuery

Answer

Cloud Spanner

Question 17

A company uses BigQuery for analytics. They have a table that is queried frequently by date range. To reduce costs, they want to ensure queries only scan the relevant partitions. They also want to improve performance for queries filtering on a specific customer_id. Which table design should they use?

Accepted Answer

Partition by date column and cluster by customer_id. Partitioning by date allows pruning irrelevant partitions; clustering on customer_id orders data within partitions for efficient filtering. Clustering alone doesn't prune partitions. Ingestion-time partitioning is based on arrival time, not logical date.

Answer

Partition by ingestion time and cluster by customer_id

Answer

Use a materialized view that filters by date and customer_id

Answer

Cluster by date column and partition by customer_id

Question 18

A company runs a nightly Dataproc batch job to process large log files. The job is idempotent and can tolerate node failures if restarted. Minimizing cost is critical. What is the most cost-effective cluster design?

Accepted Answer

Use preemptible instances for all nodes and enable automatic restart. Preemptible instances cost about 80% less than standard instances, making them the most cost-effective choice for fault-tolerant, idempotent batch jobs. Enabling automatic restart ensures that if a preemptible instance is terminated (which can happen at any time), Dataproc will automatically recreate it, maintaining cluster capacity without manual intervention. This design minimizes cost while preserving the job's ability to complete despite node failures.

Answer

Use standard instances with autoscaling based on YARN memory

Answer

Use all preemptible instances and configure the cluster to delete after the job completes

Answer

Use a single-node cluster with a high-memory machine type

Question 19

A company runs a batch ETL pipeline on Cloud Dataproc. During peak hours, the job takes longer than expected. The pipeline reads from Cloud Storage, transforms data, and writes to BigQuery. What is the most cost-effective way to improve performance without redesigning the pipeline?

Accepted Answer

Add a secondary worker group using preemptible VMs and increase the number of workers.. Adding a secondary worker group with preemptible VMs is the most cost-effective way to improve performance because it allows you to scale out the cluster horizontally with compute instances that are significantly cheaper (up to 80% discount) than regular VMs. This directly addresses the bottleneck of processing capacity during peak hours without requiring any pipeline redesign, as Cloud Dataproc can automatically distribute work across additional workers.

Answer

Enable local SSDs on all worker nodes.

Answer

Increase the master node's machine type to n1-highmem-32.

Answer

Use Cloud Composer to schedule the job with a higher priority.

Question 20

You are designing a Dataflow pipeline to process streaming data. The pipeline may encounter malformed records. You need to handle these errors without failing the entire pipeline and store the bad records for later analysis. What is the best practice?

Accepted Answer

Use a dead letter sink to write malformed records to a separate Pub/Sub topic or GCS location.. Dead letter sinks are a common pattern: route erroneous records to a separate output (e.g., Pub/Sub topic or GCS) for later investigation. Writing to BigQuery using Storage Write API with error handling is good, but for malformed records you want to isolate them. Raising exceptions would fail the pipeline. Logging only loses the data.

Answer

Catch the exception and log it, then continue processing.

Answer

Write all records to BigQuery using the Storage Write API and handle errors in the write operation.

Answer

Raise an exception in the DoFn to stop the pipeline for manual intervention.