Designing & Planning

Architecture design, solution planning, and requirements analysis

1

Set 1

20 questions • ~25 min

Start Set 1
2

Set 2

20 questions • ~25 min

Start Set 2
3

Set 3

20 questions • ~25 min

Start Set 3
4

Set 4

20 questions • ~25 min

Start Set 4
5

Set 5

9 questions • ~25 min

Start Set 5

All Questions

89 questions • Flexible time

Start All

Quick Drill

15 questions • ~15 min

Quick Drill
Choose Different Domain

Sample GCP PCA questions with explanations

These six general GCP PCA questions come from the full practice bank; they are not filtered to the selected domain. Review the answers and explanations for cross-domain practice, then choose a focused set above.

Question 1 General

You're building a RAG (Retrieval-Augmented Generation) system. What component stores and retrieves relevant document chunks based on semantic similarity?

  • ✓ Vector database (like Vertex AI Vector Search or AlloyDB with pgvector).
  • ○ Cloud SQL relational tables.
  • ○ Cloud Storage buckets.
  • ○ Pub/Sub topics.
Answer: A. Vector database (like Vertex AI Vector Search or AlloyDB with pgvector).
Vector databases store document embeddings and enable semantic similarity search, which is essential for RAG systems to retrieve relevant context for LLM prompts.
Question 2 General Choose 2

Your Cloud Run service connects to Firestore. You are seeing high latency on the 'Cold Start'. Which TWO GenAI/App optimizations can reduce perceived latency for the user?

  • ○ Turn on 'CPU Boost' in Cloud Run.
  • ✓ Use Min-Instances > 0.
  • ✓ Streaming responses (stream=True) for LLM output.
  • ○ Use a larger container image.
  • ○ Increase the timeout.
Answer: B. Use Min-Instances > 0. | C. Streaming responses (stream=True) for LLM output.
Min-instances (B) keeps the container warm, eliminating the infrastructure cold start. Streaming responses (C) begins sending text tokens to the user immediately as they generate, rather than waiting for the full completion, drastically improving 'Time to First Token' (perceived latency).
Question 3 General

You are archiving regulatory data. It must be kept for 7 years. It will likely never be accessed again unless there is an audit. Which storage class offers the lowest price for this use case?

  • ✓ Archive Storage.
  • ○ Coldline Storage.
  • ○ Nearline Storage.
  • ○ Standard Storage.
Answer: A. Archive Storage.
Archive Storage is the lowest cost tier for storage ($0.0012/GB/month), designed for data accessed less than once a year. It has the highest retrieval cost, but for 'likely never accessed', it is the cheapest choice.
Question 4 General Choose 2

You're designing a microservices architecture for an e-commerce platform. Services need to communicate asynchronously, handle traffic spikes during sales events, and ensure messages are not lost. Which TWO services should be part of your design?

  • ✓ Pub/Sub for asynchronous messaging.
  • ○ HTTP REST APIs with Cloud Run.
  • ✓ Cloud Tasks for guaranteed task execution.
  • ○ Cloud Storage for message passing.
  • ○ Memorystore for temporary message storage.
Answer: A. Pub/Sub for asynchronous messaging. | C. Cloud Tasks for guaranteed task execution.
Pub/Sub provides scalable, asynchronous messaging that can handle massive traffic spikes and ensures at-least-once delivery. Cloud Tasks provides guaranteed task execution with retry logic and rate limiting. While HTTP REST APIs are useful for synchronous communication, they don't provide the async, guaranteed delivery needed.
Question 5 General Choose 2

You're optimizing inference costs for a deployed model. Which TWO techniques help reduce costs?

  • ✓ Model quantization to reduce memory and compute requirements.
  • ✓ Batch prediction for non-real-time workloads.
  • ○ Use largest GPU available regardless of need.
  • ○ Disable autoscaling.
  • ○ Increase replica count permanently.
Answer: A. Model quantization to reduce memory and compute requirements. | B. Batch prediction for non-real-time workloads.
Quantization reduces model size and compute needs. Batch prediction is cheaper for non-real-time use cases than online endpoints.
Question 6 General

Finance wants a detailed breakdown of costs by service, project, and SKU for the past year. What's the best data source for custom cost analysis?

  • ✓ Export billing data to BigQuery and write SQL queries for custom reports.
  • ○ Use the Cloud Console billing page only.
  • ○ Request invoices from Google.
  • ○ Estimate costs manually.
Answer: A. Export billing data to BigQuery and write SQL queries for custom reports.
Billing export to BigQuery provides granular, queryable data including every line item, SKU, label, and cost. SQL enables custom analysis, dashboards, and reports. Console (B) has limited analysis capabilities. Invoices (C) are summaries. Manual estimation (D) is inaccurate.

Open the full GCP PCA question-and-answer guide