PCD Designing highly scalable, available, and reliable cloud-native applications Practice Test 5 — 15 Questions

Question 1

A mobile gaming company is launching a real-time multiplayer game. Player matchmaking data and game session state are stored in Cloud Firestore in Native mode. The game is expected to have sudden spikes in traffic, with millions of concurrent players during peak hours. The team wants to ensure the application remains highly available and can scale to handle the load without manual intervention. They also want to minimize latency for players worldwide. What should they do?

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

Accepted Answer

Deploy the game backend on Cloud Run, set a high concurrency value, and use Firestore's real-time listeners for state synchronization.. Cloud Run is a fully managed serverless platform that scales automatically in response to traffic, making it ideal for sudden spikes. Setting a high concurrency value allows each instance to handle multiple requests, improving efficiency and reducing cost. Firestore, as a globally distributed NoSQL database, provides low-latency access and real-time synchronization. Together, they form a scalable, highly available backend that requires minimal operational effort, directly addressing the game's requirements.

Answer

Use Cloud Run to host the game backend, enable CPU always allocated, and set concurrency to 1 to ensure each request gets a dedicated instance.

Answer

Deploy the game backend on a regional Google Kubernetes Engine (GKE) cluster with a Horizontal Pod Autoscaler based on CPU utilization.

Answer

Use Cloud Run to host the game backend, configure it to allow unauthenticated invocations, and set a maximum instance limit to control cost.

Question 2

A retail company runs an order-processing service on Cloud Run (fully managed) in us-central1. The service is stateless and writes to Firestore. During flash sales, traffic spikes 20x within seconds and users report HTTP 503 responses from the service. The team has already set the maximum number of instances to 100 and allotted sufficient CPU. What should they do to eliminate the 503 errors during the spike?

Accepted Answer

Increase the maximum number of instances and set a minimum instance count greater than zero to keep warm capacity available.. Cloud Run rejects requests with 503 when no instance can accept them, which happens when the revision's maximum instance count is reached during a sharp burst. Keeping a minimum instance count removes cold-start latency and preserves warm capacity, and increasing the maximum ceiling allows the autoscaler to add the instances needed for the 20x spike, so requests are served instead of rejected.

Answer

Raise the container concurrency setting to its maximum value so each instance accepts more simultaneous requests.

Answer

Deploy the service to a second region and use a global external Application Load Balancer with failover.

Answer

Configure a serverless network endpoint group (NEG) in front of the service and attach it to an external Application Load Balancer.

Question 3

A healthcare company is designing a patient records system on Google Cloud. The system must be highly available across regions and provide low-latency access to records for users in North America and Europe. The data must be strongly consistent and survive a regional outage. The team plans to use Cloud Spanner. Which TWO configurations should they implement? (Choose two.)

Accepted Answer

Configure the application to use the Cloud Spanner API with session pooling and retry logic for aborted transactions.. A multi-region Cloud Spanner configuration ensures data is replicated across regions, providing strong consistency and survival of a regional outage. Session pooling and retry logic are necessary for efficient and reliable application interaction with Spanner, especially under high load.

Answer

Enable Cloud Spanner backup and restore to a different region and schedule hourly backups.

Answer

Use a single-region Cloud Spanner instance in us-central1 and enable read replicas in Europe.

Answer

Deploy the application in a single region and use Cloud Interconnect to connect to the other region for data access.

Question 4

A media streaming company is designing a new application on Google Cloud. The application must be resilient to a full region failure and must continue serving users with minimal disruption. The company wants to use managed services and avoid managing infrastructure. Which TWO strategies should they implement? (Choose two.)

Accepted Answer

Deploy the application on Cloud Run in multiple regions and use Global External Application Load Balancer to route users to the closest healthy region.. Deploying Cloud Run across multiple regions with a Global External Application Load Balancer provides automatic failover and low-latency routing. Firestore in Native mode with a multi-region location provides a managed, globally replicated database for application state. Together, they enable a resilient, serverless architecture that survives a region failure.

Answer

Store user session data in a multi-region Cloud Storage bucket and have the application read and write directly to it.

Answer

Use Cloud SQL with a cross-region replica and manually promote the replica during a regional outage.

Answer

Deploy the application on a single Compute Engine instance with a global load balancer and a managed instance group in one region.

Question 5

A retail company runs a product catalog API on GKE Autopilot across three zones in one region. During a zonal outage, the API stayed available, but the team discovered that a StatefulSet backing a search index used a single ReadWriteOnce PersistentVolume and could not recover on another node. They now want the stateful workload to survive a zonal failure with minimal data loss and automatic recovery. (Choose two.)

Accepted Answer

Replace the PersistentVolume with a Regional persistent disk and ensure the StatefulSet pods are spread across zones with topology spread constraints.. Surviving a zonal failure with stateful workloads requires storage that is not bound to a single zone and scheduling that keeps replicas distributed. Regional persistent disks replicate synchronously across two zones and can be reattached on surviving nodes, while topology spread constraints and node auto-provisioning let the StatefulSet reschedule pods automatically. A PodDisruptionBudget and extra replicas address voluntary disruptions, not zonal outages.

Answer

Use a Zonal persistent disk with a higher IOPS tier and rely on the GKE node auto-repair feature to rebuild the node in the failed zone.

Answer

Configure the StatefulSet with a PodDisruptionBudget and increase the replica count to three.

Answer

Store the search index in a Multi-Zone bucket in Cloud Storage and mount it into the pods with the Cloud Storage FUSE CSI driver.

Question 6

A healthcare company runs a patient portal on Cloud Run (fully managed) behind a global external HTTP(S) load balancer. The portal must remain available if an entire Google Cloud region becomes unavailable, and the company wants to minimize recovery time. The container image is stored in Artifact Registry. Which configuration should they implement?

Accepted Answer

Deploy the Cloud Run service to multiple regions, add each regional service as a backend to a single global external HTTP(S) load balancer with a serverless network endpoint group, and rely on health-check-based failover.. To survive a full regional outage, the Cloud Run service must run in at least two regions and be fronted by a global external HTTP(S) load balancer that can route to healthy backends. Serverless network endpoint groups connect regional Cloud Run services to the load balancer, enabling automatic failover. Caching, DNS weighting, and in-region scaling do not provide the required regional resilience.

Answer

Deploy the Cloud Run service to multiple regions and use Cloud DNS with a weighted routing policy pointing to each regional Cloud Run URL.

Answer

Deploy the Cloud Run service in a single region but increase the maximum instance count and enable CPU always allocated to improve resilience.

Answer

Deploy the Cloud Run service in a single region and configure Cloud CDN with a long cache TTL to serve requests if the region fails.

Question 7

A company is deploying a global application on Google Cloud that uses Cloud Spanner for its database. They need to ensure that the application remains available even if an entire Google Cloud region fails. Which Spanner configuration should they use?

Accepted Answer

Multi-region configuration with two read-write regions. Cloud Spanner multi-region configurations replicate data across multiple regions, providing higher availability and the ability to survive a regional outage. A configuration with two read-write regions allows Spanner to automatically fail over and continue serving reads and writes if one region fails. This is the recommended approach for applications requiring global availability and regional fault tolerance.

Answer

Multi-region configuration with one read-write region and two read-only regions

Answer

Regional configuration with three read-write replicas

Answer

Regional configuration with two read-write replicas and one witness replica

Question 8

A financial services company is designing a highly available and scalable application on Google Cloud. The application consists of a frontend web service and a backend API, both deployed on Cloud Run. The backend API connects to a Cloud SQL for PostgreSQL instance. The company needs to ensure that the application can survive a zonal outage and continue to serve traffic without manual intervention. They also want to minimize downtime during database maintenance. Which TWO strategies should they implement? (Choose two.)

Accepted Answer

Configure the Cloud SQL instance with high availability (HA) and a failover replica in a different zone.. To survive a zonal outage and minimize database maintenance downtime, the database must have a standby in another zone. Cloud SQL high availability provides automatic failover to a standby replica. To survive a regional outage and distribute traffic globally, a global external HTTP(S) load balancer can route to Cloud Run services in multiple regions. Together, these strategies ensure the application remains available during infrastructure failures and maintenance events.

Answer

Configure the Cloud Run services to use a custom domain with Cloud DNS and set up health checks.

Answer

Enable Cloud SQL automatic backups and point-in-time recovery to restore the database in case of failure.

Answer

Deploy the Cloud Run services with a minimum instance count of 1 to ensure they are always running.

Question 9

A team runs a Node.js API on Cloud Run behind an external Application Load Balancer. During peak traffic, some request logs show '503 Service Unavailable' returned by the load balancer while Cloud Run instances report healthy startup probes. Autoscaling is set to the default maximum of 1000 instances and each instance handles 80 concurrent requests. The team wants to eliminate these 503s without over-provisioning. What should they do?

Accepted Answer

Increase the Cloud Run maximum instance count and set the container concurrency to a value matching the application's measured throughput, then retest under load.. Cloud Run scales horizontally, but the default maximum instance count and the per-instance concurrency setting determine how many simultaneous requests the service can accept. When the request queue fills, the load balancer observes 503 responses. Raising the maximum instance count and aligning concurrency with measured application throughput lets the service absorb bursts while keeping resource usage proportional to real demand.

Answer

Configure a Serverless Network Endpoint Group (NEG) backend with connection draining timeout increased and set a Cloud Armor rate-limiting rule on the backend service.

Answer

Switch the load balancer to a regional external Application Load Balancer and add a second region for failover.

Answer

Enable Cloud CDN on the load balancer backend and set a long cache TTL for all API responses to reduce origin load.

Question 10

A logistics company runs an order-tracking API on Cloud Run that reads and writes session state to Memorystore for Redis. During peak season, traffic spikes 10x and users report intermittent 500 errors. Cloud Run request concurrency is set to 80, and the Redis instance is a Basic tier 5 GB instance. The team must improve reliability with minimal application changes. What should they do?

Accepted Answer

Upgrade the Memorystore instance to a Standard tier HA instance with a replica and configure the Cloud Run service to use a Serverless VPC Access connector with the same region.. The failures come from a single-node Basic tier Memorystore instance that cannot survive a node failure or heavy load. Moving to a Standard tier HA instance with automatic replica failover removes that single point of failure, and using a Serverless VPC Access connector preserves private connectivity from Cloud Run. Because no application logic changes are needed, this meets the minimal-change constraint while improving reliability under peak traffic.

Answer

Migrate session state from Memorystore to a regional Cloud Storage bucket and mount the bucket using Cloud Run volume mounts.

Answer

Increase the Cloud Run maximum number of instances to 1000 and set the container concurrency to 250 to absorb the traffic spike.

Answer

Enable Cloud CDN in front of the Cloud Run service and set a cache TTL of 60 seconds for all API responses.

Question 11

A financial analytics team runs a containerized Python service on Google Kubernetes Engine (GKE) Autopilot. The service receives bursts of up to 40,000 requests per minute, but traffic drops to near zero overnight. The team wants the pods to scale out within seconds of a burst without over-provisioning nodes. They also want to avoid paying for idle nodes. Which scaling configuration best meets these requirements?

Accepted Answer

Deploy the service on GKE Autopilot and configure the Horizontal Pod Autoscaler (HPA) using a Pub/Sub or request-rate based external metric, allowing Autopilot to provision nodes on demand.. The requirement pairs rapid scale-out with no cost for idle capacity. Autopilot handles node lifecycle automatically, removing nodes when pods are gone, while an external request-rate metric lets HPA add replicas ahead of CPU saturation. Together they give the fast burst response and the overnight scale-to-low behavior the team needs.

Answer

Use a GKE Standard cluster with a node pool of preemptible VMs and a Vertical Pod Autoscaler (VPA) in recommendation mode.

Answer

Configure the Horizontal Pod Autoscaler (HPA) with a custom metric from Cloud Monitoring and set a high target value so pods scale aggressively.

Answer

Enable the Horizontal Pod Autoscaler (HPA) on CPU utilization and configure a cluster autoscaler with a minimum node count equal to the peak workload.

Question 12

A company is designing a highly available application on Google Cloud using Cloud Run. They want to ensure that the application can handle traffic spikes and remain available even if a single zone fails. Which two strategies should they implement? (Choose two.)

Accepted Answer

Deploy the Cloud Run service in multiple regions.. To achieve high availability and handle traffic spikes in Cloud Run, deploying the service in multiple regions provides geographic redundancy and resilience against regional failures. Configuring a minimum number of instances ensures that the service has warm instances to handle sudden traffic increases and avoids cold starts. Together, these strategies enhance availability and scalability.

Answer

Enable Cloud CDN for the Cloud Run service.

Answer

Use a single region with multiple zones.

Answer

Set the maximum number of instances to a high value.

Question 13

A media company stores user-generated videos in a Cloud Storage bucket. They need to generate thumbnails for every uploaded video. The processing takes 30 seconds and requires 2 GB of memory. The number of uploads varies unpredictably from 0 to 1000 per hour. They want a fully managed, event-driven solution that scales automatically and minimizes cost. What should they do?

Accepted Answer

Deploy a Cloud Run service that is triggered by Eventarc on the google.cloud.storage.object.v1.finalized event, and configure the service to handle retries idempotently.. Cloud Run with Eventarc provides a fully managed, event-driven solution that scales automatically and integrates directly with Cloud Storage events. The google.cloud.storage.object.v1.finalized event is triggered when an object is finalized. Cloud Run can handle the 30-second processing and 2 GB memory requirement, and it scales to zero when idle, minimizing cost. Idempotent handling ensures that duplicate events (due to at-least-once delivery) do not cause redundant processing.

Answer

Use a Cloud Function triggered by a Pub/Sub topic that receives messages when a video is uploaded, and set the function's timeout to 10 minutes.

Answer

Create a Compute Engine instance with a startup script that polls the bucket every minute for new videos and processes them.

Answer

Use a Cloud Function triggered by the object.finalize event on the bucket, with 2 GB memory and a 60-second timeout.

Question 14

A financial analytics platform stores time-series tick data in Cloud Bigtable. The application reads a single row key repeatedly to serve a live dashboard and also performs large scans over contiguous key ranges for nightly reports. Users report that dashboard reads have become slow as the table grew. The table uses a single column family and row keys based on a monotonically increasing timestamp. What should you change to improve read performance for the dashboard?

Accepted Answer

Redesign the row key to avoid a monotonically increasing prefix, for example by salting or hashing the timestamp.. Bigtable performance depends heavily on row-key distribution across tablets. A monotonic timestamp prefix causes all recent writes and the corresponding reads to land on one tablet, producing a hotspot as data grows. Salting or hashing the prefix distributes rows across tablets, which improves both the single-row dashboard read and the range scans used by nightly reports.

Answer

Add a second column family for the dashboard data and read only that family.

Answer

Enable a replication cluster in a second zone and route dashboard reads to the replica.

Answer

Increase the number of nodes in the Bigtable cluster to add more read throughput.

Question 15

A media company stores millions of user-uploaded video thumbnails in a Cloud Storage bucket. The application must serve these thumbnails to a global audience with low latency and high availability while keeping costs predictable. The team wants a design that does not depend on a single region. (Choose two.)

Accepted Answer

Front the bucket with Cloud CDN by using a global external Application Load Balancer with a backend bucket.. A multi-region Standard bucket gives geo-distributed durability and availability for the objects, while Cloud CDN behind a global external Application Load Balancer caches thumbnails at edge locations to cut latency for repeat views. Together they deliver global performance and resilience without manual replication or exposing objects publicly.

Answer

Enable Object Versioning on the bucket and set a lifecycle rule to delete noncurrent versions after 30 days.

Answer

Create a Cloud Storage transfer job to replicate the bucket to a second region daily.

Answer

Set a uniform bucket-level access policy and grant the allUsers member the Storage Object Viewer role.