PDE Practice Test 47 — 15 Questions

Question 1

Your company runs a real-time recommendation system for a popular e-commerce website using a machine learning model deployed on Vertex AI Endpoints. The model takes user features and product catalog data as input and returns top-10 product recommendations. The system uses a feature store to serve user embeddings and product embeddings. Recently, the recommender team retrained the model with a new algorithm and deployed it as a new version. Since the deployment, the latency for recommendation requests has increased from 100ms to 500ms on average, exceeding the 200ms SLO. The model accuracy is acceptable, and there are no errors. The endpoint uses an n1-standard-8 machine with a single GPU. The new model is larger but still fits on the GPU. You investigate and find that the GPU utilization remains low (<20%), but CPU utilization is high (90%). What should you do to reduce latency while maintaining accuracy?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Increase the number of replicas (nodes) to parallelize the CPU-bound preprocessing work.. The high CPU utilization (90%) with low GPU utilization (<20%) indicates that the bottleneck is CPU-bound preprocessing, not GPU inference. Increasing the number of replicas (nodes) distributes the CPU preprocessing load across multiple instances, reducing per-request latency without affecting model accuracy. This directly addresses the root cause while keeping the existing GPU resources.

Answer

Upgrade the machine type to one with more GPU memory (e.g., n1-standard-8 with a larger GPU) to reduce model inference time.

Answer

Change the batch size in the model serving code to process multiple requests together, improving GPU utilization.

Answer

Offload preprocessing to a dedicated Cloud Run service that runs asynchronously and returns precomputed feature vectors.

Question 2

Your company wants to analyze real-time user clickstream data from a website. The data arrives as JSON messages via an HTTP endpoint. The pipeline should be able to handle spikes in traffic, provide low-latency insights, and store the raw data in a data lake for historical analysis. Which Google Cloud service should you use to ingest and process the streaming data?

Accepted Answer

Cloud Pub/Sub combined with Dataflow. Cloud Pub/Sub is the correct ingestion service because it provides a highly scalable, fully managed message queue that can handle traffic spikes by decoupling producers from consumers. Dataflow (Apache Beam) then processes the streaming data with low latency, supports exactly-once semantics, and can write raw data to a data lake like Cloud Storage for historical analysis. This combination meets all requirements: spike handling, low-latency insights, and raw data storage.

Answer

Cloud Dataproc

Answer

Cloud Functions

Answer

Cloud IoT Core

Question 3

A company uses Cloud Composer to orchestrate Dataproc and BigQuery jobs. They need to implement retry logic for transient failures. Which THREE features can help?

Accepted Answer

DAG retry_delay. Cloud Composer (Apache Airflow) allows setting `retry_delay` at the DAG level to define the time delay between task retries. This is a native Airflow feature that helps handle transient failures by automatically retrying failed tasks after a specified delay, reducing manual intervention.

Answer

Dataflow pipeline retries

Answer

Cloud Composer high availability

Question 4

An e-commerce company runs a daily batch pipeline that processes clickstream data from Cloud Storage using Cloud Dataproc with Spark. The pipeline includes a join between a large fact table and a small dimension table. The dimension table is stored in Cloud Storage as a CSV file. The join is slow due to shuffling. The data engineer considers broadcasting the dimension table. However, the dimension table is updated daily and the pipeline reads the latest version. What is the best approach to implement this optimization?

Accepted Answer

Use DataFrame.join with broadcast hint on the dimension DataFrame. Broadcasting the small dimension table using the broadcast hint (e.g., `broadcast(dimensionDF)`) forces Spark to replicate the dimension data to all executor nodes, eliminating the need for a shuffle during the join. This is ideal when the dimension table is small enough to fit in executor memory, and since the pipeline reads the latest CSV daily, the broadcast will automatically use the updated data without additional code changes.

Answer

Read the fact table and dimension table into separate DataFrames and use standard join

Answer

Read the dimension table as an RDD and collect as a map, then use map-side join

Answer

Increase the spark.sql.autoBroadcastJoinThreshold to a large value

Question 5

A company wants to use Pub/Sub Lite to reduce costs for a high-throughput, low-latency streaming pipeline. However, they have a requirement to retain messages for up to 7 days for reprocessing. Which Pub/Sub Lite configuration supports this retention?

Accepted Answer

Set the retention duration on the Pub/Sub Lite topic to 7 days. Pub/Sub Lite topics are the only entity where retention duration is configured; messages are retained in the topic's storage for the specified duration, allowing subscribers to replay messages within that window. Setting the retention duration to 7 days on the topic ensures messages are available for reprocessing for up to 7 days, meeting the requirement.

Answer

Set the retention duration on the Pub/Sub Lite subscription to 7 days

Answer

Enable exactly-once delivery on the Pub/Sub Lite topic to retain messages for 7 days

Answer

Use a Pub/Sub Lite reservation with 7-day retention

Question 6

Your team wants to continuously monitor a deployed model's performance in production. They need to detect when the model's predictions become unreliable due to changes in the real world (e.g., new customer behavior). Which Vertex AI service should they use?

Accepted Answer

Vertex AI Model Monitoring. Vertex AI Model Monitoring is the correct choice because it is specifically designed to continuously track a deployed model's prediction quality over time, detecting issues like data drift, feature drift, and prediction skew that indicate the model's reliability is degrading due to changes in the real world. It automatically compares incoming prediction data against a baseline training dataset and alerts when statistical distributions shift beyond configurable thresholds, enabling proactive retraining or intervention.

Answer

Vertex AI Explainable AI

Answer

Vertex AI Experiments

Answer

Vertex AI Prediction

Question 7

Refer to the exhibit. A Dataflow streaming pipeline subscribes to this Pub/Sub subscription. The pipeline occasionally takes more than 10 seconds to process a message. Which behavior will occur?

Accepted Answer

The message will be redelivered after 10 seconds if not acknowledged.. Pub/Sub delivery requires an acknowledgment within the configurable `ackDeadlineSeconds` (default 10 seconds). If the pipeline takes longer than the ack deadline to process a message, Pub/Sub considers the message unacknowledged and redelivers it. This is the standard behavior for at-least-once delivery in Google Cloud Pub/Sub.

Answer

The message will be sent to the dead letter topic immediately.

Answer

The message will be retried with exponential backoff as per retry policy.

Answer

The message will be dropped after 10 seconds due to expiration policy.

Question 8

Which TWO statements are correct about designing a data pipeline using Cloud Dataflow for processing unbounded data?

Accepted Answer

Watermarks are used to measure the progress of event time.. Watermarks in Cloud Dataflow measure the progress of event time, indicating when all data up to a certain timestamp is expected to have arrived. This allows the pipeline to handle late-arriving data and determine when to close windows for unbounded data streams.

Answer

Triggers can only emit results at the end of a window.

Answer

Dataflow guarantees exactly-once processing for streaming pipelines.

Answer

Fixed windows are always based on processing time.

Question 9

A company runs a Dataflow streaming pipeline that reads from Cloud Pub/Sub and writes to BigQuery. The pipeline uses a side input that is a large lookup table (10 GB) stored in Cloud Storage. The side input is updated hourly. The pipeline experiences high latency and OOM errors on workers. What is the best approach to resolve this?

Accepted Answer

Use a Cloud Bigtable table as a side input via a RichSDF.. Using a Cloud Bigtable table as a side input via a RichSDF (Rich Splittable DoFn) allows the pipeline to perform point lookups on the large (10 GB) lookup table without loading it entirely into worker memory. This avoids OOM errors and reduces latency by leveraging Bigtable's low-latency, scalable key-value storage, which is ideal for high-throughput streaming pipelines that require frequent, random access to a large, frequently updated dataset.

Answer

Use a side input from a PCollection and broadcast it.

Answer

Increase the number of workers to distribute the side input.

Answer

Increase the worker memory to 16 GB per worker.

Question 10

A data engineer needs to process large CSV files (hundreds of GB) stored in Cloud Storage using Spark on a Dataproc cluster. The job performs a series of transformations and aggregations. Which configuration is most cost-effective and operationally efficient?

Accepted Answer

Use a cluster with a standard master node and 10 preemptible worker nodes (n1-standard-4).. Preemptible workers are significantly cheaper (about 80% discount) and ideal for batch processing of large CSV files where fault tolerance is built into Spark via RDD lineage. Using standard nodes for the master ensures cluster stability, while preemptible workers handle the distributed transformations and aggregations cost-effectively. This configuration balances cost and operational efficiency for ephemeral, fault-tolerant workloads.

Answer

Use a cluster with 10 high-memory (n1-highmem-8) VMs as workers to improve shuffle performance.

Answer

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

Answer

Use a cluster with 10 standard (n1-standard-4) VMs as master and worker nodes, all non-preemptible.

Question 11

A financial services company uses Cloud Composer to orchestrate daily batch jobs. One job extracts data from MongoDB to Cloud Storage, then loads into BigQuery, and finally runs a Dataflow pipeline for aggregations. The Dataflow job fails intermittently. They want to automatically restart only the failed Dataflow job without re-running the earlier extraction and load. Which Airflow operator configuration should they use?

Accepted Answer

Set retries=2 on the Dataflow operator. Setting retries=2 on the Dataflow operator instructs Airflow to automatically restart only that specific task upon failure, without affecting upstream tasks (MongoDB extraction, BigQuery load). This isolates the retry to the Dataflow job, preserving the earlier completed work and avoiding redundant data movement.

Answer

Implement a SlaMiss sensor

Answer

Use a DAG with depends_on_past=True

Answer

Set trigger_rule='one_success' for downstream tasks

Question 12

A machine learning team wants to deploy a new model version for canary testing, where only 5% of traffic is routed to the new version. Which Vertex AI endpoint configuration supports this?

Accepted Answer

Configure the endpoint with traffic split: 95% to old version, 5% to new version.. Vertex AI endpoints natively support traffic splitting, allowing you to route a specified percentage of requests to different model versions deployed on the same endpoint. By configuring a traffic split of 95% to the old version and 5% to the new version, you can perform canary testing without additional infrastructure or client-side logic. This is the correct and simplest approach within Vertex AI.

Answer

Have the client application randomly select which model to call with 5% probability.

Answer

Deploy the new version to a separate endpoint and direct 5% of users via a load balancer.

Answer

Use an A/B testing framework outside of Vertex AI to compare results.

Question 13

A company has deployed a machine learning model to AI Platform Prediction. The model uses a custom container with a TensorFlow SavedModel. After deployment, the prediction latency is higher than expected. Which action is most likely to reduce latency without significantly impacting model accuracy?

Accepted Answer

Enable XLA (Accelerated Linear Algebra) compilation on model loading.. Enabling XLA (Accelerated Linear Algebra) compilation on model loading optimizes the TensorFlow computation graph by fusing operations and reducing runtime overhead, which directly lowers prediction latency without altering model weights or architecture. XLA works by compiling the graph into efficient machine code at load time, improving execution speed while preserving the original model accuracy.

Answer

Convert the model to TensorFlow Lite and use a smaller model.

Answer

Increase the number of prediction nodes in the AI Platform Prediction cluster.

Answer

Apply quantization to the model weights to reduce size.

Question 14

A retailer wants to use machine learning to predict customer churn based on transaction history and demographic data. The dataset has 500 features, many of which are correlated. The data is highly imbalanced: only 2% churn. They need to deploy a model that provides feature importance and is interpretable. Which model type should they use in BigQuery ML?

Accepted Answer

Boosted tree classifier. Boosted tree models (like XGBoost) handle imbalanced data well, provide feature importance, and offer interpretability. Deep Neural Networks are less interpretable. Logistic regression is interpretable but may not capture complex patterns. AutoML Tables is powerful but less interpretable and may cost more.

Answer

Logistic regression

Answer

AutoML Tables model

Answer

Deep Neural Network (DNN) classifier

Question 15

An organization needs to prevent data exfiltration from BigQuery by ensuring all traffic to BigQuery APIs goes through VPC boundaries and is restricted to a specific service perimeter. Which Google Cloud security control should they use?

Accepted Answer

VPC Service Controls. VPC Service Controls (D) is the correct answer because it allows you to define a service perimeter around BigQuery APIs, ensuring that all traffic to BigQuery must originate from within the defined VPC boundaries. This prevents data exfiltration by blocking unauthorized access from outside the perimeter, even if valid credentials are used. It works by enforcing context-aware access policies at the Google Cloud network edge, not at the application layer.

Answer

Access Transparency

Answer

IAM conditions on BigQuery roles

Answer

Cloud Armor