Courseiva

Google Cloud Digital Leader (GCDL) — Questions 451–525

848 questions total · 12 pages · All types, answers revealed

Page 6

Page 7 of 12

Page 8
451
MCQhard

A DevOps engineer wants to audit all actions performed by Google personnel on their customer data stored in Cloud Storage. They need to review logs that show access by Google employees and the reason for access. Which logging feature should they enable?

A.Cloud Audit Logs
B.Cloud Logging
C.Access Transparency
D.VPC Flow Logs
AnswerC

Access Transparency is the correct feature because it provides real-time logs of every action taken by Google personnel (and certain subprocessors) when accessing customer content, including reads, writes, and administrative operations. It complements Cloud Audit Logs by covering the 'Google-side' of the shared responsibility model, which no other logging option addresses. It must be explicitly enabled on the organization, folder, or project, and it works only for a defined set of Google Cloud services.

Why this answer

Access Transparency provides logs specifically showing when Google personnel access customer data, including the reason and justification for access. It is designed for auditing Google employee actions on customer data, which is exactly the requirement.

Exam trap

The trap is confusing Cloud Audit Logs with Access Transparency: candidates pick Audit Logs because they think it covers all access, but only Access Transparency logs Google personnel access with justification.

How to eliminate wrong answers

Option A is wrong because Cloud Audit Logs record actions by users and services in your project, not access by Google personnel to your data. Option B is wrong because Cloud Logging is a general log management service; it does not specifically capture Google employee access with justification. Option D is wrong because VPC Flow Logs capture network traffic metadata for VPC subnets, not personnel access to data.

452
MCQmedium

A team uses multiple cloud services and wants to deploy all resources — VPCs, Cloud SQL databases, GKE clusters, and IAM roles — using a declarative, open-source infrastructure-as-code tool that works across multiple cloud providers. Which tool integrates natively with Google Cloud for this purpose?

A.Cloud Deployment Manager — Google's native IaC service.
B.Terraform
C.Cloud Build — it builds and deploys application code.
D.Ansible — it automates server configuration management.
AnswerB

HashiCorp Terraform is an open-source, declarative Infrastructure as Code tool that defines resources in HashiCorp Configuration Language (HCL) and maintains a state file to plan and apply changes. Its Google Cloud provider comprehensively covers all GCP resources, while the same HCL configuration can also manage AWS, Azure, and other clouds, making it the industry standard for multi-cloud infrastructure provisioning. Terraform's resource graph and state-driven planning enable safe, incremental updates that are difficult to achieve with configuration management or CI/CD tools.

Why this answer

Terraform is the correct choice because it is a declarative, open-source infrastructure-as-code tool that supports multiple cloud providers, including Google Cloud, through its provider plugin architecture. It allows you to manage VPCs, Cloud SQL databases, GKE clusters, and IAM roles using HashiCorp Configuration Language (HCL) and integrates natively with Google Cloud via the google provider.

Exam trap

The trap here is that candidates often confuse Cloud Deployment Manager (a Google-native, proprietary tool) with a multi-cloud solution, or mistake Cloud Build (a CI/CD tool) for an IaC tool, when the question explicitly requires an open-source, multi-cloud declarative IaC tool.

How to eliminate wrong answers

Option A is wrong because Cloud Deployment Manager is Google's native IaC service, but it is not open-source and only works within Google Cloud, not across multiple cloud providers. Option C is wrong because Cloud Build is a CI/CD service for building and deploying application code, not a declarative IaC tool for managing cloud resources like VPCs or databases. Option D is wrong because Ansible is a configuration management and automation tool focused on server provisioning and application deployment, not a declarative IaC tool for managing cloud infrastructure across providers.

453
MCQhard

A company is running a PostgreSQL database on Cloud SQL and needs to ensure high availability with automatic failover in the event of a zone failure. Which configuration should they use?

A.Enable regional persistent disk and configure a standby instance in a different zone.
B.Use point-in-time recovery.
C.Configure connection pooling.
D.Set up a cross-region read replica.
AnswerA

Cloud SQL's high availability configuration uses a regional persistent disk, which synchronously replicates data between zones, combined with a standby instance in a different zone. In the event of a zonal outage, Cloud SQL automatically detects the failure and promotes the standby, typically within tens of seconds. Because replication is synchronous, the RPO is effectively zero, ensuring no committed transactions are lost. This is the correct choice for an availability requirement within a single region.

Why this answer

Enabling regional persistent disk allows the primary and standby Cloud SQL instances to share the same underlying storage across zones. When a zone failure occurs, the standby instance in a different zone automatically takes over with no data loss, providing high availability with automatic failover.

Exam trap

Google Cloud often tests the distinction between high availability (automatic failover within a region) and disaster recovery (manual or cross-region failover), leading candidates to mistakenly choose cross-region read replicas for HA scenarios.

How to eliminate wrong answers

Option B is wrong because point-in-time recovery (PITR) is a backup and restore feature that allows recovering to a specific timestamp, not a mechanism for automatic failover or high availability. Option C is wrong because connection pooling manages database connections to improve performance and reduce overhead, but it does not provide failover or zone redundancy. Option D is wrong because a cross-region read replica is designed for read scaling and disaster recovery across regions, not for automatic failover within the same region; it requires manual promotion and does not provide automatic failover for the primary instance.

454
MCQeasy

A small marketing agency currently runs its customer relationship management (CRM) application on a physical server in its office closet. The server is five years old, and the agency's IT consultant has warned that a disk failure could take the CRM offline for days. The agency wants to move the CRM to Google Cloud so that it can scale capacity up or down as campaign volumes change, without buying new hardware. Which cloud computing benefit does this scenario primarily illustrate?

A.Vendor lock-in, allowing the agency to depend on a single provider for all services
B.Elasticity, allowing resources to scale automatically with demand
C.Colocation, allowing the server to be housed in a third-party data center
D.High availability, guaranteeing the CRM will never experience an outage
AnswerB

Elasticity means the agency can provision more compute or storage during busy campaign periods and release it afterward, paying only for what it uses. This directly solves the problem of fixed physical hardware that cannot grow or shrink, and it eliminates the multi-day downtime risk tied to a single aging server in an office closet.

Why this answer

The agency's core problem is fixed on-premises capacity that cannot flex with campaign demand and depends on aging hardware. Elasticity lets compute and storage resources expand during peak periods and contract afterward, with consumption-based billing. This removes the need to purchase new servers and directly addresses the scaling requirement described in the scenario.

Exam trap

The trap here is assuming that any cloud benefit involving reliability or scaling automatically maps to high availability, when the scenario is really about matching capacity to fluctuating demand.

455
Multi-Selectmedium

A company has multiple projects and wants to organize them by environment (dev, test, prod) and by team (engineering, marketing, finance). They also need to apply IAM policies that affect all projects in a given environment. Which TWO steps should they take?

Select 2 answers
A.Create separate billing accounts for each environment
B.Use organization policies at the project level
C.Apply IAM policies at the folder level
D.Use labels to group projects by environment
E.Create folders for each environment (dev, test, prod) and place projects in the appropriate folder
AnswersC, E

Folders are hierarchical nodes in the resource hierarchy, and any IAM policy attached to a folder is inherited by all projects and resources within that folder. This lets you bind environment-wide roles, such as DevTeam or Viewer, once rather than repeating them on each project, giving consistent access across every project in the environment. It is the correct pattern because it uses native inheritance rather than per-project duplication.

Why this answer

Using folders to group projects by environment allows inheritance of IAM policies. Labels are used for cost attribution and filtering, not for policy enforcement.

456
MCQmedium

A security engineer notices that a Compute Engine instance is running a VM with a public IP that should not be accessible from the internet. They want to ensure this configuration is prevented by default for all future projects in the organization. What should they do?

A.Set an IAM policy to deny compute.instances.create with public IP
B.Define an Organization Policy with the constraint compute.vmExternalIpAccess
C.Create a VPC firewall rule to deny all traffic from the internet to the VM
D.Use Cloud Security Scanner to identify and remediate
AnswerB

The Organization Policy constraint compute.vmExternalIpAccess is a list constraint that can be applied at the organization, folder, or project level to explicitly block the creation of Compute Engine VMs with external IP addresses. This is a preventive, infrastructure-level guardrail enforced by the Resource Manager at request time, regardless of the IAM roles held by the caller. It can be configured to deny all external IPs or to allow only a set of specific IP ranges, and it propagates hierarchically to all child resources.

Why this answer

Organization Policies in Google Cloud allow you to set constraints at the organization, folder, or project level to enforce security controls. The `compute.vmExternalIpAccess` constraint specifically prevents VMs from being created with external IP addresses, ensuring that no future Compute Engine instances in the organization can have public IPs by default. This is a preventive control that applies to all new VM creations, unlike IAM policies or firewall rules which are more granular or reactive.

Exam trap

The trap here is that candidates often confuse IAM policies with Organization Policies, thinking that IAM can restrict resource configurations (like public IPs) when it only controls who can perform actions, not the attributes of the resources created.

How to eliminate wrong answers

Option A is wrong because IAM policies control who can perform actions (like `compute.instances.create`), but they cannot restrict the configuration of a resource (such as whether a public IP is assigned) — IAM does not support conditional constraints on resource attributes like external IP assignment. Option C is wrong because a VPC firewall rule can block traffic to the VM, but it does not prevent the VM from having a public IP address; the VM would still be reachable from the internet if the firewall rule is misconfigured or not applied, and it does not enforce a default policy for future projects. Option D is wrong because Cloud Security Scanner is a tool for finding vulnerabilities in web applications (like XSS or CSRF), not for enforcing organizational policies on VM public IP assignment; it is a detective control, not a preventive one.

457
MCQhard

A healthcare company runs its critical application on Google Cloud. The application uses Cloud SQL for patient records, Cloud Storage for medical images, and Pub/Sub for data ingestion. The security team requires that all data at rest be encrypted with a key that is managed and rotated by their on-premises HSM. They also need to ensure that any potential data exfiltration is immediately detected and prevented. Recently, a vulnerability scan revealed that a Cloud SQL instance had a public IP. The team wants to enforce that no Cloud SQL instance can be created with a public IP across the entire organization. Additionally, they need to implement a solution to monitor and alert on any suspicious activity, such as a large download from Cloud Storage. They have a limited budget and cannot afford complex custom solutions. Which combination of Google Cloud services should they use to meet these requirements?

A.Use CMEK with Cloud KMS for encryption, set an Organization Policy to restrict public IPs on Cloud SQL, and configure Cloud Audit Logs with alerting via Cloud Monitoring to detect data exfiltration.
B.Use Cloud External Key Manager (EKM) for encryption, define an Organization Policy constraint to prohibit public IPs on Cloud SQL, deploy Security Command Center with Event Threat Detection to monitor for data exfiltration, and implement VPC Service Controls to limit data access.
C.Use default encryption with Google-managed keys, set an IAM condition to deny public IP on Cloud SQL, and configure Cloud Data Loss Prevention to detect sensitive data exfiltration.
D.Use Cloud HSM for encryption, create a VPC firewall rule to block all incoming traffic to Cloud SQL, and use Cloud Armor to protect against data exfiltration.
AnswerB

This is the correct answer because Cloud External Key Manager (EKM) integrates with an on-premises HSM, enabling the healthcare company to maintain control of encryption keys outside Google Cloud, satisfying the key management requirement. An Organization Policy constraint that prohibits public IPs on Cloud SQL is the proper resource-level enforcement mechanism. Security Command Center with Event Threat Detection monitors network and API activity for signs of data exfiltration, while VPC Service Controls creates a security perimeter that prevents data from being copied or transferred out of authorized services, addressing both detection and prevention.

Why this answer

Cloud External Key Manager (EKM) allows you to use an external key management system (on-premises HSM) for encrypting data at rest in Google Cloud services like Cloud SQL, Cloud Storage, and Pub/Sub. The Organization Policy constraint `constraints/sql.restrictPublicIp` can enforce that no Cloud SQL instance is created with a public IP. Security Command Center with Event Threat Detection provides out-of-the-box monitoring and alerting for suspicious activities like large downloads from Cloud Storage, while VPC Service Controls adds a data exfiltration prevention layer by restricting data movement outside a defined service perimeter.

Exam trap

Google Cloud often tests the distinction between key management options (CMEK vs. EKM vs. Cloud HSM) and the difference between detection (Cloud Audit Logs, Event Threat Detection) and prevention (VPC Service Controls), leading candidates to choose a solution that only detects but does not prevent data exfiltration.

How to eliminate wrong answers

Option A is wrong because CMEK with Cloud KMS uses keys managed within Google Cloud, not an on-premises HSM, and Cloud Audit Logs with Cloud Monitoring alone cannot prevent data exfiltration—they only provide logging and alerting, not active prevention. Option C is wrong because default encryption uses Google-managed keys, not customer-managed keys from an on-premises HSM, and IAM conditions cannot enforce a restriction on Cloud SQL public IPs at the organization level (that requires an Organization Policy). Option D is wrong because Cloud HSM is a Google-managed HSM service, not an on-premises HSM, and VPC firewall rules cannot block public IP assignment on Cloud SQL (they control network traffic, not resource configuration), while Cloud Armor is a web application firewall, not a data exfiltration detection or prevention tool.

458
MCQeasy

Which Google Cloud service provides a fully managed, serverless data warehouse for petabyte-scale analytics with SQL?

A.Cloud SQL
B.BigQuery
C.Dataproc
D.Dataflow
AnswerB

BigQuery is Google Cloud's serverless, highly scalable, SQL-based data warehouse. It automatically manages infrastructure and scales compute and storage independently, using a columnar storage format and a distributed query engine (Dremel) to run analytics on petabytes of data. With a pay-per-query pricing model and no clusters to provision, BigQuery is the definitive choice for a fully managed data warehouse on Google Cloud.

Why this answer

BigQuery is Google Cloud's fully managed, serverless data warehouse. It supports SQL queries at petabyte scale with no infrastructure to manage. Cloud SQL is for OLTP, Dataproc is for Hadoop/Spark, and Dataflow is for stream/batch processing.

459
MCQeasy

A startup wants to launch a new mobile app globally. They expect user traffic to be unpredictable and want to only pay for the compute resources they use. Which cloud benefit BEST addresses this need?

A.Global Reach
B.Agility
C.Pay-as-you-go pricing
D.Scalability
AnswerC

Pay-as-you-go pricing is a usage-based billing model where cloud customers pay per unit of measured consumption (e.g., vCPU-hours, GB-months, or network requests) with no upfront capital expenditures or minimum commitments. For a startup with limited budget and uncertain initial user volume, this model directly matches the requirement of paying only for resources actually used, because the bill scales with demand and can be near-zero when the app has no traffic. It avoids the waste of idle capacity that is inherent in fixed-cost models like reserved instances or on-premises hardware. This is the precise reason it is the correct choice for the described cost-efficiency need.

Why this answer

The pay-as-you-go model allows startups to avoid large upfront capital expenditure and only pay for actual usage, which is ideal for unpredictable workloads.

460
MCQeasy

An organization wants to reduce latency for users in Europe. They plan to deploy their application in a Google Cloud region located in Europe. Which region should they choose?

A.us-central1
B.asia-east1
C.australia-southeast1
D.europe-west1
AnswerD

europe-west1 is located in St. Ghislain, Belgium, placing it in close geographic proximity to major European population centers such as London, Paris, and Frankfurt. This region benefits from Google's low-latency European backbone and multiple availability zones, enabling round-trip times in the tens of milliseconds for users across Western Europe. Deploying workloads here minimizes network hops and intercontinental cable crossings, directly addressing the goal of reducing latency for users in Europe.

Why this answer

europe-west1 is a Google Cloud region physically located in Belgium, so deploying there places compute and storage close to European users, reducing network round-trip latency. The other options are regions on different continents, which would add transatlantic or transpacific latency for European users.

Exam trap

The trap is overthinking region selection — candidates may look for a specific city or assume a US region is a safe default, but the question simply tests whether you know which listed region is in Europe.

How to eliminate wrong answers

Option A is wrong because us-central1 is in Iowa, USA, so European users would incur transatlantic latency. Option B is wrong because asia-east1 is in Taiwan, which is even farther from Europe and would add significant latency. Option C is wrong because australia-southeast1 is in Sydney, Australia, which is on the opposite side of the globe from European users.

461
Multi-Selecthard

Which TWO of the following cloud characteristics directly enable a business to innovate faster than using traditional IT?

Select 2 answers
A.Global infrastructure that allows launching in new regions quickly
B.Vendor lock-in for long-term contracts
C.Capital expenditure model for budgeting
D.Customizable hardware configurations for optimal performance
E.Self-service provisioning of resources in minutes
AnswersA, E

Cloud providers operate globally distributed regions and edge locations, enabling organizations to deploy workloads nearer to end users with low latency. This geographic reach supports compliance with data residency requirements and disaster recovery strategies. It accelerates international expansion by removing the need for physical data center construction, which traditionally takes months or years.

Why this answer

A global infrastructure with multiple regions and edge locations allows businesses to deploy applications and services in new geographic areas rapidly, reducing time-to-market compared to building or leasing physical data centers. Option E is correct because self-service provisioning enables developers to spin up resources like virtual machines, databases, or containers in minutes via APIs or console, eliminating the weeks-long procurement and setup cycles of traditional IT, thus accelerating experimentation and iteration.

Exam trap

Google Cloud often tests the misconception that capital expenditure (CapEx) is more predictable and thus faster for innovation, but the trap is that CapEx actually introduces procurement delays and financial friction, whereas cloud's operational expenditure (OpEx) model enables rapid, on-demand scaling without upfront investment.

462
MCQhard

A company wants to receive near-real-time notifications when spending in a project exceeds 50%, 75%, and 100% of a monthly budget of $50,000. They also want to automatically disable billing for the project if spending exceeds $60,000. Which configuration should they use?

A.Use the Cost Management dashboard to monitor spending and disable billing manually.
B.Create a budget alert at 100% only and manually disable billing when alerted.
C.Create a budget with thresholds at 50%, 75%, 100% and set the 'disable billing' action on the budget.
D.Create a budget with thresholds at 50%, 75%, 100% and set up a Cloud Function that listens to Pub/Sub messages from the budget to automatically disable billing when a 120% threshold is exceeded.
AnswerD

This is the correct architecture: Cloud Billing budgets can send threshold-alert messages to a Pub/Sub topic when spend crosses defined percentages (e.g., 50%, 75%, 100%, 120%). A Cloud Function subscribes to that topic, parses the budget notification payload, and if the cost exceeds the intended final threshold (120%), it invokes the Cloud Billing API to disable the billing account. This provides near-real-time, automated enforcement without manual steps, while the earlier thresholds serve as warnings that can trigger other notifications or actions.

Why this answer

Budgets can be set with multiple threshold rules for notifications. However, automatically disabling billing requires a separate automation (e.g., Cloud Functions triggered by Pub/Sub) because budgets do not perform actions beyond notifications and Pub/Sub messages.

463
MCQhard

A multinational corporation needs to establish a private, low-latency connection between their on-premises data center and Google Cloud. The connection must be consistent, reliable, and support at least 10 Gbps throughput. Which solution should they use?

A.Cloud NAT with static IP addresses
B.Dedicated Interconnect
C.Partner Interconnect via a service provider
D.Cloud VPN using IPsec tunnels over the public internet
AnswerB

Dedicated Interconnect is the correct choice because it establishes one or more direct, physical connections between your on-premises network and Google's edge network using colocation facilities. These private links offer high throughput (10 Gbps or 100 Gbps per connection) and low, consistent latency because traffic does not traverse the public internet. It is ideal for latency-sensitive, bandwidth-heavy workloads that demand a private, reliable, and dedicated network path between a multinational corporation and GCP.

Why this answer

Dedicated Interconnect provides a direct, private physical connection between an on-premises network and Google Cloud, offering consistent, reliable, low-latency connectivity with throughput options of 10 Gbps or 100 Gbps per link. This meets the requirement for a private, low-latency connection with at least 10 Gbps throughput, as it bypasses the public internet entirely and provides a Service Level Agreement (SLA) for uptime and performance.

Exam trap

Google Cloud often tests the distinction between Dedicated and Partner Interconnect, where candidates mistakenly choose Partner Interconnect for higher throughput, but Dedicated Interconnect is the only option that provides direct, private physical links at 10 Gbps or 100 Gbps without a service provider intermediary.

How to eliminate wrong answers

Option A is wrong because Cloud NAT is used for outbound internet access from private instances and does not provide a private, low-latency connection between on-premises and Google Cloud; it is a network address translation service, not a connectivity solution. Option C is wrong because Partner Interconnect relies on a supported service provider and typically offers lower throughput options (e.g., 50 Mbps to 10 Gbps) and may introduce additional latency or dependency on the provider's network, whereas the requirement specifies a direct, private connection with at least 10 Gbps throughput, which Dedicated Interconnect fulfills without a third-party intermediary. Option D is wrong because Cloud VPN uses IPsec tunnels over the public internet, which introduces variable latency, potential packet loss, and lower throughput (typically up to 3 Gbps per tunnel), and does not provide the consistent, reliable, low-latency performance required for a 10 Gbps connection.

464
MCQmedium

A security administrator needs to grant a developer the minimum permissions to create and delete Cloud Storage buckets in a specific project, but NOT allow them to modify the contents of those buckets (e.g., upload or delete objects). Which IAM role should they assign?

A.Storage Admin (roles/storage.admin)
B.Storage Object Admin (roles/storage.objectAdmin)
C.Custom role with storage.buckets.* permissions
D.Project Editor (roles/editor)
AnswerC

A custom role with storage.buckets.* permissions is the correct choice because it allows you to define a precise set of bucket-level operations—such as storage.buckets.create, storage.buckets.get, storage.buckets.update, and storage.buckets.delete—while explicitly excluding all object-related permissions (e.g., storage.objects.*). This granularity enforces least privilege by ensuring the developer can manage the bucket lifecycle but cannot view, upload, or delete any objects stored within those buckets. Unlike predefined roles that couple bucket and object access, this custom role minimizes the attack surface and aligns with security best practices for infrastructure management.

Why this answer

Storage Admin (roles/storage.admin) grants full control over buckets and objects, which is too permissive. The correct approach is to create a custom role with only storage.buckets.* permissions, or use predefined roles like Storage Object Admin? Actually, Storage Object Admin allows object management. The question requires bucket management only.

A custom role with storage.buckets.create and storage.buckets.delete (and maybe storage.buckets.get) is necessary; predefined roles do not separate bucket and object permissions perfectly. But among standard roles, 'Storage Admin' is too broad; there is no predefined role that only allows bucket management without object access. So the answer must be a custom role.

465
MCQmedium

A company wants to use machine learning to analyze customer reviews without building and training models from scratch. They need a pre-trained model that can classify sentiment. Which Google Cloud service should they use?

A.Cloud Natural Language API
B.Vertex AI
C.AutoML Natural Language
D.Dialogflow
AnswerA

Cloud Natural Language API is the correct choice because it offers pre-trained models that can immediately perform sentiment analysis on customer text via a simple REST call. There is no need to build, train, or deploy a custom model, and the API returns sentiment scores and magnitude out of the box. This makes it the fastest and most cost-effective approach for analyzing customer feedback without prior ML expertise.

Why this answer

The Cloud Natural Language API is a pre-trained, fully managed service that exposes sentiment analysis (among other NLP features like entity and syntax analysis) through a simple REST/RPC call, so no model training or ML expertise is required. It directly satisfies the requirement for a pre-trained sentiment classification model. Vertex AI and AutoML require you to build/train or at least manage a model, and Dialogflow is for conversational agents, not sentiment classification.

Exam trap

GCDL often tests the distinction between pre-trained ML APIs (ready to call, no training) and custom model platforms (Vertex AI, AutoML) that require labeled data and training — candidates confuse 'using ML' with 'building ML'.

How to eliminate wrong answers

Option B is wrong because Vertex AI is a full ML platform for building, training, and deploying custom models — it does not provide a ready-made sentiment classifier out of the box. Option C is wrong because AutoML Natural Language requires you to supply labeled training data and train a custom model, which contradicts the 'without building and training models' requirement. Option D is wrong because Dialogflow is a conversational AI/chatbot platform for intent and entity recognition in dialogue flows, not a general-purpose sentiment analysis API.

466
Multi-Selecthard

A company wants to reduce costs for its batch processing jobs that run nightly on Compute Engine. The jobs are fault-tolerant and can be interrupted. They are considering using preemptible VMs. Which THREE statements about preemptible VMs are true?

Select 3 answers
A.Preemptible VMs can be migrated to regular VMs if preemption occurs.
B.Preemptible VMs do not offer live migration.
C.Preemptible VMs provide the same SLA as standard VMs.
D.Preemptible VMs can run for up to 24 hours before they may be terminated.
E.Preemptible VMs are significantly cheaper than standard VMs.
AnswersB, D, E

Unlike standard Compute Engine VMs, which benefit from live migration during infrastructure maintenance, preemptible VMs are not live-migrated. If the underlying host needs maintenance or Google reclaims capacity, the VM is immediately terminated within 30 seconds of the preemption notice. This means the VM's state is lost unless you have explicitly saved it elsewhere.

Why this answer

Preemptible VMs can be terminated at any time within 24 hours (typical max 24h). They are significantly cheaper than regular VMs. They cannot be migrated to regular VMs; you must recreate them.

They do not offer live migration. They are suitable for fault-tolerant batch jobs.

467
MCQmedium

A solutions architect is explaining why using managed cloud database services (like Cloud SQL or Cloud Spanner) is preferable to running a database on a self-managed virtual machine in most cases. What is the primary operational advantage of managed database services over self-managed databases on VMs?

A.Managed databases are always significantly cheaper than self-managed databases on VMs
B.Managed database services automate operational tasks like backups, patching, HA failover, and scaling — freeing engineering teams to focus on application development rather than database administration
C.Managed databases guarantee better query performance than self-managed databases for all workload types
D.Managed databases provide stronger data encryption than self-managed databases on VMs
AnswerB

This is the core value proposition of managed databases. The cloud provider handles: automated daily backups with point-in-time recovery, OS and database software patching, automatic failover for high availability, and storage scaling. Engineering teams avoid the specialized DBA work required for self-managed databases.

Why this answer

Managed database services like Cloud SQL and Cloud Spanner abstract away the operational overhead of database administration. They automate critical tasks such as automated backups, patch management, high-availability failover, and horizontal scaling, which allows engineering teams to focus on application logic rather than managing database servers, replication, or storage.

Exam trap

The GCDL exam often tests the misconception that managed services are always cheaper or always faster, when the primary advantage is operational automation and reduced administrative burden, not cost or performance guarantees.

How to eliminate wrong answers

Option A is wrong because managed databases are not always significantly cheaper; they often have higher per-hour costs than self-managed VMs, though they can reduce total cost of ownership by eliminating administrative labor and infrastructure overhead. Option C is wrong because managed databases do not guarantee better query performance for all workload types; performance depends on instance size, query optimization, and workload characteristics, and self-managed databases can be tuned more aggressively for specific use cases. Option D is wrong because both managed and self-managed databases can implement strong encryption (e.g., AES-256 at rest, TLS 1.3 in transit); encryption is a configuration choice, not an inherent advantage of managed services.

468
MCQeasy

A developer is troubleshooting a slow response from a Cloud Run service. Which Google Cloud service can they use to trace requests across microservices?

A.Cloud Profiler
B.Cloud Trace
C.Cloud Logging
D.Cloud Debugger
AnswerB

Cloud Trace collects and aggregates latency data (trace spans) from distributed services, allowing developers to view the end-to-end journey of a single request, including delays in each service, RPC calls, and external API calls. It provides a waterfall view, which directly helps diagnose slow responses by identifying the bottleneck span. This makes it the correct tool for troubleshooting response latency.

Why this answer

Cloud Trace is the correct service because it is specifically designed for distributed tracing, collecting latency data from applications and displaying it in a trace timeline. It can trace requests as they propagate across multiple microservices, including Cloud Run services, by using trace context propagation headers (e.g., `X-Cloud-Trace-Context`). This allows the developer to identify bottlenecks and slow components in a request path.

Exam trap

The trap here is that candidates often confuse Cloud Trace with Cloud Logging, thinking that log aggregation alone can reconstruct request paths, but Cloud Trace is the only service that provides distributed tracing with explicit span context propagation across microservices.

How to eliminate wrong answers

Option A is wrong because Cloud Profiler is a statistical, low-overhead profiler that identifies which code paths consume the most CPU or memory, not a tool for tracing individual request flows across microservices. Option C is wrong because Cloud Logging aggregates and stores log entries but does not provide end-to-end request tracing or visualize the path of a single request across services. Option D is wrong because Cloud Debugger allows you to inspect the state of a running application at a specific code point without stopping it, but it does not trace request propagation or measure latency across services.

469
MCQeasy

A company wants to replace its VPN-based remote access with a solution that grants access based on user identity, device security status, and context (e.g., location, IP). Which Google Cloud service should they use?

A.BeyondCorp Enterprise
B.Cloud Armor
C.Identity-Aware Proxy (IAP)
D.Cloud VPN
AnswerA

BeyondCorp Enterprise is Google's zero-trust access solution that fully replaces a VPN-based remote access model. It grants users access to specific applications based on verified identity, device trust, and context (such as location and security posture) rather than opening up network-level connectivity. This eliminates the implicit trust and lateral movement risks inherent in traditional VPNs, while also integrating with Google Cloud's security capabilities like DLP and threat detection.

Why this answer

BeyondCorp Enterprise is Google Cloud's zero-trust access platform that grants access based on user identity, device security posture, and contextual signals (location, IP, time), replacing traditional VPN-based remote access. It builds on Identity-Aware Proxy (IAP) and adds device trust, continuous authorization, and threat/data protection. This matches the requirement for identity- and context-based access without a VPN.

Exam trap

The trap is picking IAP because it is the technical enforcement mechanism — but the question asks for the service that provides identity, device, and context-based access, which is the broader BeyondCorp Enterprise offering.

How to eliminate wrong answers

Option B is wrong because Cloud Armor is a WAF and DDoS protection service for external HTTP(S) load balancers, not a zero-trust remote access solution. Option C is wrong because IAP is a component of BeyondCorp Enterprise — it provides identity-based access to apps and VMs, but does not by itself deliver the full device-posture and context-aware policy engine that BeyondCorp Enterprise provides. Option D is wrong because Cloud VPN is a traditional VPN solution, which is exactly what the company wants to replace.

470
Multi-Selecthard

A company is building a microservices architecture on Google Kubernetes Engine (GKE). They need to expose services externally with HTTPS, distribute traffic across the cluster, and protect against DDoS attacks. Which THREE Google Cloud services should they combine? (Choose THREE)

Select 3 answers
A.Cloud DNS
B.VPC firewall rules
C.Cloud CDN
D.Cloud Armor
E.Cloud Load Balancing
AnswersC, D, E

Cloud CDN caches content at Google's global edge points of presence, absorbing volumetric traffic before it reaches your GKE backends. This directly satisfies the DDoS protection constraint: attack requests terminate at edge locations rather than consuming cluster resources, while also reducing origin load and latency for externally exposed HTTPS services.

Why this answer

Cloud Load Balancing (E) is correct because it provides the external HTTPS load balancer that fronts the GKE services, terminates TLS, and distributes incoming traffic across the cluster's nodes and pods. Cloud Armor (D) is correct because it attaches to the external load balancer's backend service to provide WAF rules and Layer 3/4/7 DDoS protection, including Google Cloud's edge-based network DDoS mitigation. Cloud CDN (C) is correct because it caches content at Google's global edge locations, reducing origin load and absorbing traffic spikes, which complements the load balancer and helps mitigate volumetric attacks.

Cloud DNS (A) is not required for this scenario since it only resolves domain names and does not distribute traffic or provide DDoS protection, and VPC firewall rules (B) operate at the network level within the VPC but do not expose services externally over HTTPS or defend against DDoS at the edge.

Exam trap

The trap is including Cloud DNS or VPC firewall rules as part of the traffic distribution and protection solution; candidates must distinguish name resolution and network-level filtering from the edge services that actually load balance, cache, and defend against DDoS.

471
MCQhard

A security engineer needs to analyze network traffic for malicious payloads and anomalies in real-time across multiple VPC networks in a project. The solution must be managed and not require deploying third-party appliances. Which service should they use?

A.Security Command Center
B.Cloud Armor
C.Cloud IDS
D.VPC Flow Logs
AnswerC

Cloud IDS is a fully managed intrusion detection service that performs deep packet inspection (DPI) on all network traffic within a VPC, including both east-west and north-south flows. It leverages the industry-leading threat signatures from Palo Alto Networks to detect malicious payloads, command-and-control communications, and other network-based attacks. By comparing packet payloads against known signatures, Cloud IDS provides real-time visibility into actual traffic content, directly meeting the requirement to analyze network traffic for threats.

Why this answer

Cloud IDS provides managed intrusion detection across VPC networks, analyzing traffic for threats like malware and anomalies.

472
MCQhard

A company is using Cloud SQL for MySQL and notices that read queries are becoming slow as the application scales. They want to offload read traffic from the primary instance to improve performance. Which Cloud SQL feature should they enable?

A.Automatic failover replicas
B.Connection pooling
C.Point-in-time recovery
D.Read replicas
AnswerD

Read replicas are asynchronous copies of the primary instance that support only read-only traffic. By routing SELECT queries to read replicas, you reduce the read load on the primary and increase overall scalability. This is the correct approach for offloading reads, though applications must be configured to use the replica endpoints, and data may experience a slight replication lag.

Why this answer

Read replicas in Cloud SQL for MySQL are designed specifically to offload read traffic from the primary instance by maintaining an asynchronous copy of the data that applications can query directly. By directing SELECT queries to one or more read replicas, the primary instance is freed up to handle write operations, improving overall throughput and reducing latency on the primary. This is the standard horizontal scaling pattern for read-heavy Cloud SQL workloads.

Exam trap

The trap here is confusing high-availability/failover replicas with read replicas — both are called 'replicas,' but only read replicas serve client read traffic, while failover replicas are passive standbys.

How to eliminate wrong answers

Option A is wrong because automatic failover replicas (high-availability configuration) exist to provide automatic failover of the primary instance in the event of a zone or instance failure, not to serve read traffic — they are a standby and do not accept client reads. Option B is wrong because connection pooling reduces the overhead of opening and closing database connections; it improves connection efficiency but does not distribute read queries away from the primary instance. Option C is wrong because point-in-time recovery is a backup/restore capability that lets you recover a database to a specific timestamp, and has nothing to do with scaling read throughput.

473
MCQmedium

A technology company runs its containerized microservices on Google Kubernetes Engine (GKE). The development team frequently pushes new container images to Container Registry, and those images are deployed to a production cluster. The security team recently discovered that a few running containers have critical vulnerabilities from outdated base images. They want to enforce a policy that only vulnerability-scanned and approved images can be deployed in the production cluster. The team uses Cloud Build for CI/CD and Container Analysis for vulnerability scanning. Which solution should they implement to meet this requirement?

A.Use Cloud Security Scanner to scan the production cluster for vulnerabilities.
B.Enable Cloud Asset Inventory to monitor image vulnerabilities across projects.
C.Configure Cloud Build to run a vulnerability scan step before pushing images to Container Registry.
D.Enable Binary Authorization with a policy that requires attestations from Container Analysis for all deployments in the production cluster.
AnswerD

Binary Authorization enforces a supply-chain policy at deployment time by requiring trusted attestations for every container image deployed to the GKE cluster. You can configure Container Analysis (e.g., on-demand scanning) to generate attestations for images that pass a vulnerability threshold, and an Attestor can be set to require those attestations. Because this is enforced by the Kubernetes admission controller in the cluster, any attempt to deploy an unverified image is blocked — providing the actual enforcement that the other options lack.

Why this answer

Binary Authorization enforces deployment-time policies that require signed attestations from trusted authorities (like Container Analysis) before an image can be deployed on GKE. By configuring a policy that mandates an attestation from Container Analysis (which performs vulnerability scanning), only images that have been scanned and approved can be deployed, directly meeting the requirement to block containers with critical vulnerabilities.

Exam trap

The trap here is that candidates confuse scanning images (which only identifies vulnerabilities) with enforcing a policy that blocks deployment of vulnerable images, leading them to choose a scanning-only option (like C) instead of the policy enforcement mechanism (Binary Authorization).

How to eliminate wrong answers

Option A is wrong because Cloud Security Scanner is designed to find web application vulnerabilities (e.g., XSS, SQLi) in App Engine, Compute Engine, and GKE services, not to enforce deployment policies or scan container images for OS-level vulnerabilities. Option B is wrong because Cloud Asset Inventory provides a historical view of cloud resources and their metadata (including vulnerability findings from Container Analysis), but it cannot enforce a policy that blocks deployments; it is a monitoring and inventory tool, not a policy enforcement mechanism. Option C is wrong because running a vulnerability scan step before pushing images to Container Registry only ensures images are scanned at build time, but it does not prevent a developer from bypassing the scan or deploying an older, unscanned image; it lacks the deployment-time enforcement that Binary Authorization provides.

474
MCQhard

An organization stores sensitive data in BigQuery. They need to restrict access to specific columns based on user role, while allowing analysis at the dataset level. Which feature should they use?

A.BigQuery row-level security
B.Column-level access control using authorized views or taxonomy policies
C.IAM roles at the dataset level with fine-grained permissions
D.Cloud Data Loss Prevention (DLP) to mask data
AnswerB

BigQuery column-level access control can be implemented via authorized views that expose only designated columns while hiding sensitive ones, or via taxonomy policies in Data Catalog that tag columns and enforce fine-grained ACLs. Authorized views allow you to share a filtered projection without granting access to the underlying table, and taxonomies with policy tags enforce column-level restrictions at query time. These mechanisms directly satisfy the need to protect sensitive data in specific columns.

Why this answer

BigQuery column-level access control, implemented through authorized views or taxonomy policies (via Data Catalog), allows restricting access to specific columns while preserving dataset-level analysis permissions. Authorized views use SQL logic to expose only permitted columns, and taxonomy policies apply fine-grained access controls at the column level without requiring separate datasets.

Exam trap

Google Cloud often tests the distinction between row-level and column-level access controls, and the trap here is that candidates confuse row-level security (which filters rows) with column-level security (which restricts columns), or mistakenly think IAM dataset-level roles can achieve fine-grained column restrictions.

How to eliminate wrong answers

Option A is wrong because BigQuery row-level security restricts access to specific rows based on filters, not columns, and does not address column-level restrictions. Option C is wrong because IAM roles at the dataset level provide coarse-grained access to entire tables or datasets, but cannot restrict access to individual columns within a table. Option D is wrong because Cloud Data Loss Prevention (DLP) is used for data discovery, classification, and masking of sensitive data, but it does not enforce persistent column-level access control for ongoing query access; it is a scanning and transformation tool, not an access control mechanism.

475
MCQmedium

A cloud team receives an alert that a critical production service's error rate has spiked. Following incident response best practices, what is the correct first priority action?

A.Identify and fix the root cause before taking any other action to ensure the fix is complete
B.Mitigate user impact immediately (e.g., rollback, traffic rerouting, scaling) while beginning parallel investigation of the root cause
C.Wait to understand the full scope of the issue and inform all stakeholders before taking any technical action
D.Escalate to senior leadership and wait for their approval before making any production changes
AnswerB

Mitigation first is the correct incident response approach. Stop the bleeding before diagnosing the cause. If a recent deployment caused the spike, roll back immediately. If it's a capacity issue, scale up. Investigation into root cause runs in parallel but mitigation is prioritized.

Why this answer

Incident response best practices prioritize reducing user impact first. In Google Cloud, this could involve rolling back a deployment via Cloud Deploy, rerouting traffic with a load balancer, or scaling up instances with Managed Instance Groups, all while a parallel investigation into the root cause begins. This aligns with the SRE principle of 'error budget' and the 'mitigate before diagnose' approach.

Exam trap

The trap here is that candidates confuse 'root cause analysis' with 'first response' — Google Cloud often tests the principle that immediate mitigation (e.g., rollback, scaling) takes precedence over diagnosis, even if the fix is temporary.

How to eliminate wrong answers

Option A is wrong because it violates the incident response principle of 'stop the bleeding' first; waiting to fix the root cause before mitigating impact prolongs user downtime and can violate SLAs. Option C is wrong because waiting to understand the full scope before taking action delays mitigation, increasing user impact and potentially breaching SLOs; parallel investigation is key. Option D is wrong because escalating for approval before acting introduces unnecessary latency; incident response requires immediate technical action to restore service, with post-incident review for leadership.

476
MCQeasy

A retail company wants to migrate its on-premises e-commerce platform to Google Cloud. The application is stateless and runs on virtual machines. The company wants to minimize operational overhead and allow the application to automatically scale based on CPU utilization. Which Google Cloud service should they use?

A.Google Kubernetes Engine (GKE)
B.Cloud Run
C.Compute Engine with managed instance groups and autoscaling
D.App Engine Standard Environment
AnswerC

Compute Engine with managed instance groups (MIGs) provides a straightforward lift-and-shift path for existing e-commerce VMs while enabling automatic scaling based on resource utilization such as CPU or memory. MIGs offer managed instance health checks, auto-healing, and autoscaling, allowing the infrastructure to adapt to traffic spikes without requiring any application refactoring. This makes it the most appropriate choice for migrating a legacy workload that needs to retain its VM-based architecture and control over the operating system.

Why this answer

The application runs on virtual machines and is stateless, making Compute Engine with managed instance groups and autoscaling the most direct fit. Managed instance groups automatically handle scaling based on CPU utilization without requiring containerization or code changes, minimizing operational overhead while preserving the existing VM-based architecture.

Exam trap

The trap here is that candidates often choose GKE or Cloud Run because they associate 'autoscaling' with Kubernetes or serverless, but the question specifies 'virtual machines' and 'minimize operational overhead,' which points to a VM-native solution like Compute Engine with MIGs, not containerization or serverless platforms.

How to eliminate wrong answers

Option A is wrong because Google Kubernetes Engine (GKE) requires containerizing the application, which adds operational overhead for managing clusters and containers, contradicting the goal of minimizing overhead. Option B is wrong because Cloud Run is a serverless platform for containerized applications that scales based on HTTP requests, not CPU utilization, and does not support virtual machines. Option D is wrong because App Engine Standard Environment is a fully managed platform for specific runtimes (e.g., Python, Java) and does not support custom virtual machines or CPU-based autoscaling for arbitrary VM images.

477
Multi-Selectmedium

A company is building a real-time analytics pipeline on Google Cloud. They need to ingest streaming data from IoT devices, process it with low latency, and then store the results for real-time querying. Which TWO services should they use? (Choose TWO.)

Select 2 answers
A.Cloud Pub/Sub
B.Cloud Storage
C.Cloud Dataflow
D.Cloud Functions
E.BigQuery
AnswersA, C

Cloud Pub/Sub provides the durable, globally scalable messaging layer that ingests streaming telemetry from IoT devices, decoupling producers from consumers. It satisfies the ingestion stage of the pipeline, buffering events until Dataflow processes them with low latency.

Why this answer

Cloud Pub/Sub (A) is correct because it is Google Cloud's fully managed, globally scalable messaging service designed for ingesting high-volume streaming data from sources such as IoT devices, decoupling producers from consumers with low-latency, at-least-once delivery. Cloud Dataflow (C) is correct because it is a fully managed Apache Beam runner that performs stream and batch processing with low latency, including windowing, aggregations, and exactly-once semantics, making it the right choice to transform the Pub/Sub stream before writing results downstream. Cloud Storage (B) is not appropriate as the primary ingestion or processing layer for real-time streaming, since it is object storage optimized for durable blobs rather than low-latency event streams.

Cloud Functions (D) is event-driven and serverless but is intended for lightweight, short-lived functions, not sustained low-latency stream processing pipelines. BigQuery (E) is a powerful analytics warehouse for real-time querying of stored results, but it is a destination/serving layer rather than the ingestion or stream-processing service the scenario requires.

Exam trap

The trap is selecting BigQuery as a processing service — it is the analytics destination, not the stream processor; candidates also confuse Cloud Functions with Dataflow for stream processing.

478
MCQeasy

An organization wants to group related projects under a common parent for policy enforcement and cost tracking. Which GCP resource hierarchy level should be used to group several projects that belong to the same business unit?

A.Organization node
B.Folder
C.Project
D.Resource
AnswerB

Folders are hierarchical containers that live beneath an Organization node and above Projects, specifically designed to group related projects. They also support nested sub-folders, enabling you to mirror your company's structure or deployment environments (e.g., dev, staging, prod). Policies and IAM roles assigned to a folder are inherited by all projects and resources inside it, making them a first-class administrative boundary for access control and isolation. This is exactly the correct mechanism for grouping related projects while preserving policy inheritance.

Why this answer

Folders are the GCP resource hierarchy level designed to group projects under a common parent, enabling policy inheritance (IAM, Org Policy) and cost aggregation for a business unit. A folder can contain projects and other folders, forming a tree under the organization node. This makes folders the correct choice for grouping several projects belonging to the same business unit.

Exam trap

GCDL often tests the confusion between organization, folder, and project levels — candidates pick the organization node for business-unit grouping when folders are the correct scoped grouping mechanism.

How to eliminate wrong answers

Option A is wrong because the organization node is the root of the hierarchy and represents the entire company, not a single business unit — using it would apply policies to all projects company-wide. Option C is wrong because a project is the resource container itself, not a grouping mechanism for multiple projects. Option D is wrong because 'resource' is a generic term for any GCP object (VMs, buckets, etc.) and is not a hierarchy level for grouping projects.

479
MCQmedium

An engineering team uses the `--preemptible` flag when creating an instance for a batch data processing job that runs nightly and can tolerate interruptions. What business transformation benefit does using the `--preemptible` flag provide?

A.Lower cost for fault-tolerant workloads, enabling more experimentation
B.Higher reliability for critical applications
C.Better performance due to dedicated hardware
D.Enhanced security through automatic patching
AnswerA

Preemptible instances are priced significantly lower than standard instances (often up to 60-80% discount), making them attractive for batch jobs, data processing, and stateless workloads that can tolerate interruptions. Because they can be terminated anytime within 24 hours, cost savings free budget for running more experiments, increasing development velocity and testing coverage.

Why this answer

The `--preemptible` flag creates a preemptible VM instance that can be terminated by Google Cloud at any time, typically after 24 hours, but at a significantly lower cost (up to 60-80% discount). For batch data processing jobs that run nightly and can tolerate interruptions, this cost savings directly enables more experimentation and iterative development because teams can run more jobs or larger datasets for the same budget, accelerating innovation and business transformation.

Exam trap

Google Cloud often tests the misconception that preemptible instances are for reliability or performance, when in fact they are a cost-optimization feature specifically for fault-tolerant, interruptible workloads.

How to eliminate wrong answers

Option B is wrong because preemptible instances are less reliable — they can be terminated at any time, making them unsuitable for critical applications that require high availability. Option C is wrong because preemptible instances do not provide dedicated hardware; they run on shared capacity that can be reclaimed, and performance may vary due to potential preemption. Option D is wrong because the `--preemptible` flag has no relation to security or automatic patching; those are managed separately through OS patch management or container image updates.

480
MCQeasy

A company wants to ensure that their customer data stored in BigQuery is encrypted at rest using customer-managed encryption keys (CMEK). Which Google Cloud service should they use to manage these keys?

A.Cloud HSM
B.Identity-Aware Proxy (IAP)
C.Cloud Key Management Service (Cloud KMS)
D.Secret Manager
AnswerC

Cloud Key Management Service (Cloud KMS) is the core service for customer-managed encryption keys (CMEK), allowing you to create, import, rotate, disable, and destroy keys that control the encryption of data in Google Cloud services like Cloud Storage, Compute Engine, and BigQuery. When you use CMEK, each service requests the key from Cloud KMS to perform envelope encryption, and Cloud KMS enforces IAM-based permissions and audit logging on every key operation. This gives customers the ability to manage the full key lifecycle, making Cloud KMS the correct choice for the stated requirement.

Why this answer

Cloud Key Management Service (Cloud KMS) allows customers to create, manage, and use encryption keys, including CMEK for BigQuery and other GCP services.

481
MCQmedium

A startup wants to launch a social media app globally. They have no existing IT infrastructure and very limited capital. The app will experience unpredictable traffic patterns, with usage expected to rapidly grow after viral campaigns. They need low latency for users across North America, Europe, and Asia. The development team is small and wants to focus on coding rather than operations. They also need to store user-generated content like images and videos. The CTO is evaluating whether to build on-premises or use cloud services. Which approach best meets their needs?

A.Deploy a single large virtual machine in one region and rely on a CDN to serve content globally.
B.Build the app on Compute Engine with managed instance groups, use Cloud CDN for global low-latency delivery, and Cloud Storage for user content.
C.Purchase and configure servers in a single colocation facility, and use a content delivery network (CDN) for static assets.
D.Use a hybrid cloud model: keep a small on-premises server for core features and burst to the cloud for extra capacity.
AnswerB

Managed instance groups give you autoscaling and automatic healing, so your app can handle unpredictable global traffic spikes by adding or removing VMs based on load. Cloud CDN accelerates delivery of static assets—like CSS, JavaScript, and media—by caching them at edge locations closer to users, while Cloud Storage provides durable, globally accessible object storage for user-generated content. This combination is fully managed, requires no upfront hardware investment, and lets you start in one region then expand to multiple regions as you grow, which is exactly what a startup needs for a global launch.

Why this answer

It leverages Google Cloud's fully managed services to meet the startup's needs: Compute Engine with managed instance groups provides auto-scaling for unpredictable traffic, Cloud CDN ensures low-latency global content delivery, and Cloud Storage offers scalable, durable storage for user-generated content. This serverless-like approach minimizes operational overhead, allowing the small team to focus on coding.

Exam trap

Google Cloud often tests the misconception that a CDN alone can solve global latency for a dynamic app, but candidates must recognize that CDNs only cache static content and do not reduce latency for dynamic requests, which require compute resources close to the user.

How to eliminate wrong answers

Option A is wrong because a single large VM in one region creates a single point of failure and cannot provide low latency across North America, Europe, and Asia; a CDN only caches static content, not dynamic app logic, so users far from that region will experience high latency. Option C is wrong because purchasing and configuring servers in a single colocation facility requires significant upfront capital and ongoing operational management, contradicting the limited capital and small team constraints; a CDN for static assets does not address dynamic request latency or auto-scaling for viral traffic spikes. Option D is wrong because a hybrid cloud model still requires maintaining on-premises servers, which incurs capital expenditure and operational overhead, and the core features running on-premises would suffer from latency for global users; it also fails to provide the fully managed, auto-scaling infrastructure needed for unpredictable growth.

482
MCQeasy

A developer accidentally commits an application's Google Cloud service account key to a public GitHub repository. The key is valid and grants access to production resources. What is the correct immediate response?

A.Delete the commit from GitHub history using git rebase; the key is safe once removed from the repository
B.Immediately revoke/delete the exposed service account key in Google Cloud IAM, review Cloud Audit Logs for unauthorized access, and generate a new key distributed through secure channels
C.Change the service account's permissions to read-only to limit the damage from potential misuse
D.Send an internal email informing the security team and wait for their guidance before taking any action
AnswerB

This is the complete correct response: (1) Revoke the key immediately to stop any ongoing unauthorized access. (2) Review Admin Activity and Data Access audit logs to determine if the key was used after exposure. (3) Issue a new key through a secure distribution channel (ideally Secret Manager, not environment variables). Time to revocation is critical.

Why this answer

The immediate priority is to invalidate the exposed credential to prevent unauthorized access to production resources. Revoking the key in Google Cloud IAM ensures it can no longer be used for authentication, while reviewing Cloud Audit Logs helps identify any potential misuse. Generating a new key and distributing it securely restores access for legitimate applications.

Exam trap

The trap here is that candidates may think removing the key from the repository (Option A) is sufficient, but they overlook that the key remains valid in Google Cloud and can still be used by anyone who already obtained it.

How to eliminate wrong answers

Option A is wrong because deleting the commit from GitHub history does not invalidate the key; anyone who already cloned or forked the repository still has access to the key, and the key remains valid in Google Cloud until explicitly revoked. Option C is wrong because changing the service account's permissions to read-only does not prevent an attacker from using the key to authenticate; the key itself is still valid and could be used for any action the service account is allowed, including reading sensitive data. Option D is wrong because waiting for guidance delays the critical step of revoking the exposed key, increasing the window of opportunity for unauthorized access; immediate action is required to contain the breach.

483
MCQhard

A company runs an e-commerce platform on Google Kubernetes Engine (GKE) using autoscaling. They have a baseline workload and occasional traffic spikes during promotions. They configured a Horizontal Pod Autoscaler (HPA) for their web application pods and a Cluster Autoscaler for the node pool. The HPA targets 70% CPU utilization. During a recent sales event, traffic exceeded expectations. The operations team observed that the HPA increased the desired number of replicas to 50, but only 20 pods were running. The remaining 30 pods were in 'Pending' status. The Cluster Autoscaler logs show repeated messages: 'no capacity to scale up node pool'. The node pool is configured with a maximum of 10 nodes, each with 4 vCPUs, and currently all 10 nodes are running. The team checked the node pool's current utilization and found that nodes are near capacity. What should the team do to ensure the application scales correctly during future events?

A.Increase the HPA target CPU utilization to 90% to reduce the number of replicas needed.
B.Reduce the pod resource requests for CPU so that more pods can fit on existing nodes.
C.Increase the maximum number of nodes in the node pool to allow more capacity.
D.Enable extra capacity by creating a second node pool with preemptible VMs.
AnswerC

Increasing the maximum number of nodes in the node pool is the correct fix because the Cluster Autoscaler has already scaled to the current maximum of 10 nodes, yet pending pods remain. By raising that ceiling, the Cluster Autoscaler can provision additional nodes to absorb the scheduler's pending pod queue, restoring capacity while preserving the CPU headroom defined by the HPA and pod resource requests. This directly addresses the root cause—the node pool's capacity ceiling—and allows both horizontal pod autoscaling and cluster autoscaling to work as intended.

Why this answer

The HPA requested 50 replicas, but only 20 could be scheduled because the existing 10 nodes are near capacity and the node pool is already at its maximum size. The Cluster Autoscaler cannot add more nodes because the node pool's maximum node limit prevents it from provisioning the needed additional capacity. Increasing the maximum number of nodes in the node pool (Option C) allows the Cluster Autoscaler to add nodes and accommodate the pending pods, enabling the HPA to scale as needed.

Exam trap

Google Cloud often tests the misconception that adjusting HPA thresholds or pod resource requests alone can solve capacity issues, when the real bottleneck is an insufficient node pool maximum, which must be increased to allow the Cluster Autoscaler to add capacity.

How to eliminate wrong answers

Option A is wrong because increasing the HPA target CPU utilization to 90% would reduce the number of replicas triggered by CPU, but the underlying capacity shortage remains; pods would still be pending if the node pool cannot grow. Option B is wrong because reducing pod CPU requests might allow more pods per node, but it does not address the node pool's hard limit of 10 nodes; once nodes are full, the Cluster Autoscaler still cannot add more nodes. Option D is wrong because creating a second node pool with preemptible VMs could provide additional capacity, but preemptible VMs can be terminated at any time (within 24 hours) and are not suitable for handling critical traffic spikes; the more direct and reliable fix is to increase the maximum node count in the existing node pool.

484
MCQmedium

An enterprise wants employees to access internal web applications securely from any location (including remote work from home) without using a VPN. Employees should only access apps they're authorized for, based on their identity and device context. Which Google Cloud service enables this zero-trust access model?

A.Cloud VPN with split tunneling for internal application access.
B.Cloud Identity-Aware Proxy (IAP)
C.Cloud Armor IP allowlist to restrict access to corporate office IP ranges.
D.Cloud Load Balancing with SSL termination.
AnswerB

Cloud Identity-Aware Proxy (IAP) is a zero-trust access layer that sits between users and applications, enforcing access policies based on the user's Google identity, OAuth2 scopes, and contextual attributes like device security status. When a request is made, IAP authenticates the principal, verifies the ID token, and checks against defined access-level conditions before proxying traffic to the backend. This eliminates the need for a VPN, as access decisions are made per request at the application layer, not based on network location.

Why this answer

Cloud Identity-Aware Proxy (IAP) is the correct service because it enforces zero-trust access by verifying a user's identity and device context before granting access to internal web applications, without requiring a VPN. It uses Google's BeyondCorp model to authenticate and authorize each request based on identity and context, allowing secure access from any location.

Exam trap

The trap here is that candidates often confuse network-level security (like VPNs or IP allowlists) with identity-aware access control, assuming that any encrypted tunnel or IP restriction satisfies zero-trust requirements, but zero-trust fundamentally requires per-request identity and context verification, not just network perimeter controls.

How to eliminate wrong answers

Option A is wrong because Cloud VPN with split tunneling still requires a VPN tunnel and does not provide identity- or device-context-based authorization; it only encrypts traffic and routes it to the internal network. Option C is wrong because Cloud Armor IP allowlisting restricts access based on source IP addresses, which fails for remote workers with dynamic IPs and does not verify user identity or device context. Option D is wrong because Cloud Load Balancing with SSL termination only handles traffic distribution and decryption, not authentication or authorization based on user identity and device posture.

485
MCQhard

An administrator wants to enforce that all API calls to a specific Cloud Storage bucket must come from a limited range of IP addresses. Which configuration should they use?

A.Cloud Armor security policy
B.Identity-Aware Proxy (IAP)
C.VPC Service Controls with an access level that includes the IP range
D.VPC firewall rules
AnswerC

VPC Service Controls with an access level that includes the IP range is correct because these perimeters restrict all API calls to protected services like Cloud Storage, and access levels apply context-aware conditions (including source IP address) at the Google Front End. When you add Cloud Storage to a service perimeter and attach an access level with your allowed CIDR range, any API request originating outside that IP range is denied, regardless of IAM permissions. This gives you perimeter-based, context-sensitive enforcement directly on the Cloud Storage API.

Why this answer

VPC Service Controls with an access level that includes the specified IP range is the correct way to enforce that API calls to a Cloud Storage bucket originate only from a limited set of IP addresses. Access levels in VPC Service Controls define conditions such as IP ranges, and when attached to a service perimeter, they restrict access to the protected service (Cloud Storage) based on those conditions. This provides centralized, identity- and context-aware enforcement at the API level.

Exam trap

GCDL often tests the confusion between network-layer controls (firewall rules, Cloud Armor) and API-layer controls (VPC Service Controls), leading candidates to choose a firewall or load balancer solution for an API-level restriction.

How to eliminate wrong answers

Option A is wrong because Cloud Armor security policies protect HTTP(S) load balancer backends and do not govern direct API calls to Cloud Storage. Option B is wrong because Identity-Aware Proxy (IAP) controls access to web applications and VMs, not to Cloud Storage API endpoints. Option D is wrong because VPC firewall rules operate at the network layer for VM traffic and do not apply to Cloud Storage API calls, which are handled by Google's API infrastructure outside the VPC.

486
MCQmedium

A company wants to proactively identify underutilized Compute Engine VMs (high provisioned capacity but low actual usage) to reduce costs. Which Google Cloud tool provides recommendations for right-sizing VMs?

A.Cloud Monitoring — set alerts for low CPU utilization.
B.Active Assist Recommender — ML-based VM rightsizing recommendations.
C.Cloud Asset Inventory — lists all VMs and their configurations.
D.Cloud Billing budgets — set spending limits to prevent overspend.
AnswerB

Active Assist Recommender is the correct service because it uses machine learning trained on near-real-time utilization metrics from Compute Engine, including CPU, memory, network, and disk I/O, to generate specific rightsizing recommendations for each VM. For example, it may suggest moving from n2-standard-8 to n2-standard-4 based on sustained low usage, and it also provides an estimated monthly savings and a confidence score for every recommendation. These recommendations are actionable directly from the console or via the Recommender API, allowing you to apply the machine type change in one click or through infrastructure-as-code. This is precisely the targeted, data-driven guidance the scenario asks for—not just raw metrics or aggregate alerts.

Why this answer

Active Assist Recommender is Google Cloud's ML-based service that analyzes historical VM utilization metrics and generates rightsizing recommendations, flagging VMs that are over-provisioned relative to actual usage. It directly addresses the goal of proactively identifying underutilized Compute Engine VMs to reduce cost. The recommendations include specific machine type suggestions and estimated savings.

Exam trap

GCDL often tests the difference between monitoring/alerting tools and recommendation engines, tempting candidates to pick Cloud Monitoring because it 'watches' utilization, when the question specifically asks for proactive rightsizing recommendations.

How to eliminate wrong answers

Option A is wrong because Cloud Monitoring alerts only notify you when CPU utilization drops below a threshold — they do not analyze usage patterns or recommend a specific right-sized machine type. Option C is wrong because Cloud Asset Inventory is a metadata inventory service that lists resources and configurations; it has no utilization analytics or recommendation engine. Option D is wrong because Cloud Billing budgets only track and alert on spend against a threshold — they do not identify which VMs are underutilized or suggest rightsizing actions.

487
Multi-Selectmedium

A company is migrating its on-premises PostgreSQL database to Google Cloud. They need a managed service that is fully compatible with PostgreSQL, offers high availability, and provides automated backups. Which TWO Google Cloud services should they consider?

Select 2 answers
A.Cloud SQL
B.Memorystore
C.AlloyDB
D.Filestore
E.Cloud Bigtable
AnswersA, C

Cloud SQL is a fully managed relational database service that supports PostgreSQL, providing automated backups, point-in-time recovery, and high availability with synchronous replication across zones. It offers native PostgreSQL compatibility, so applications can migrate without any changes to their SQL or data access layer. With managed maintenance, scaling, and built-in security features like IAM integration and encryption at rest, Cloud SQL is the pragmatic default choice for most PostgreSQL migrations to Google Cloud.

Why this answer

Cloud SQL (A) is correct because it is a fully managed relational database service that supports PostgreSQL with full compatibility, offers high availability through regional configurations with automatic failover, and provides automated backups and point-in-time recovery. AlloyDB (C) is also correct because it is a fully managed PostgreSQL-compatible database service designed for high availability with a multi-node architecture and automated backups, while offering enhanced performance for demanding workloads. Memorystore (B) is incorrect because it is a managed in-memory data store for Redis and Memcached, not a PostgreSQL-compatible relational database.

Filestore (D) is incorrect because it is a managed network file storage service (NFS) for file-based workloads, not a database service. Cloud Bigtable (E) is incorrect because it is a fully managed NoSQL wide-column database, not PostgreSQL-compatible.

Exam trap

GCDL often tests the confusion between Cloud SQL and AlloyDB, where candidates might think AlloyDB is not PostgreSQL-compatible or that Cloud SQL is not fully managed, when in fact both are valid options for PostgreSQL migration.

488
Matchingmedium

Match each Google Cloud security concept to its description.

Drag a concept onto its matching description — or click a concept then click the description.

Concepts
Matches

Identity and Access Management – fine-grained access control

Key Management Service for encryption keys

DDoS protection and web application firewall

Perimeter security to prevent data exfiltration

Centralized vulnerability and threat monitoring

Why these pairings

The correct matches are Cloud IAM (access control), Cloud KMS (key management), Cloud Security Command Center (threat detection), and Cloud DLP (data loss prevention). Common confusions include mixing up IAM with Security Command Center or KMS with Cloud Armor.

489
Multi-Selecthard

A company is planning to migrate a legacy monolithic Linux application to Google Cloud. They want to minimize changes initially but have the flexibility to modernize later. Which three approaches should they consider?

Select 3 answers
A.Use Cloud Run for stateless containers
B.Replatform to App Engine Flexible Environment
C.Use Migrate for Anthos
D.Refactor to microservices on GKE
E.Rehost on Compute Engine using a custom image
AnswersA, C, E

If the legacy application is already containerized or can be containerized with minimal effort, Cloud Run executes stateless containers in a fully managed, serverless environment; it auto-scales to zero, requires no infrastructure administration, and only necessitates the container to listen on a port. Because it does not impose a specific application framework or force a rewrite, it qualifies as a minimal-change migration for a monolithic app that does not depend on persistent local state.

Why this answer

Option A is correct because Cloud Run runs stateless containers, so the legacy Linux application can be packaged into a container with minimal code changes and then modernized incrementally later. Option C is correct because Migrate for Anthos converts existing VM-based workloads into containers running on GKE, allowing a lift-and-shift migration with a path to modernization. Option E is correct because rehosting on Compute Engine with a custom image is the classic lift-and-shift approach, requiring almost no application changes while preserving future modernization options.

Option B is not marked correct because App Engine Flexible requires replatforming the application to fit its runtime and deployment model, which is more change than the scenario wants initially. Option D is not marked correct because refactoring to microservices on GKE is a major architectural change, not a minimal-change migration.

Exam trap

Google Cloud often tests the distinction between 'minimal changes' (rehosting or container lift-and-shift) and 'modernization' (refactoring or replatforming), leading candidates to incorrectly select options that require significant code or architecture changes.

490
Multi-Selecthard

A team is designing a CI/CD pipeline for a microservices application. They want to automatically build container images from source code, store them securely, and deploy to GKE. Which THREE services should they include? (Choose three.)

Select 3 answers
A.Cloud Build
B.Cloud Storage
C.Cloud Run
D.Artifact Registry
E.GKE
AnswersA, D, E

Cloud Build is the fully managed CI/CD service that compiles your source code, runs tests, and builds the container image from a Dockerfile or build config. It executes the build steps defined in your pipeline, generating OCI-compliant images that are then pushed to a registry. This makes Cloud Build the correct core engine for the CI/CD pipeline.

Why this answer

Cloud Build (A) is correct because it is Google Cloud's managed CI/CD service that compiles source code and builds container images, and it integrates natively with GKE and Artifact Registry for automated pipelines. Artifact Registry (D) is correct because it securely stores and manages container images (and other artifacts) with IAM-based access control and vulnerability scanning, serving as the image repository the pipeline pushes to and GKE pulls from. GKE (E) is correct because it is the target runtime for deploying the containerized microservices, and it can pull images directly from Artifact Registry within the same project.

Cloud Storage (B) is not the right choice because it is object storage for blobs, not a container image registry with the tagging and vulnerability features needed here. Cloud Run (C) is not correct because it is a serverless container platform, not the GKE deployment target specified in the scenario.

Exam trap

GCDL often tests the confusion between Cloud Run and GKE as deployment targets, or between Cloud Storage and Artifact Registry for storing images; candidates must match the services to the specific requirements of building, storing, and deploying to GKE.

491
Multi-Selectmedium

A company wants to protect sensitive data stored in Cloud Storage from being downloaded by users outside their organization. They also need to prevent data from being copied to external projects. Which TWO services should they use? (Choose two.)

Select 2 answers
A.IAM conditions with access levels
B.Cloud DLP
C.VPC Service Controls
D.Cloud KMS
E.Cloud Armor
AnswersA, C

IAM conditions with access levels restrict Cloud Storage access based on request context, such as network origin or device state, blocking downloads by users outside the organisation. This satisfies the requirement to prevent external parties from downloading sensitive objects.

Why this answer

IAM conditions with access levels (A) are correct because they let you enforce context-aware access rules on Cloud Storage, such as allowing access only when the request originates from a trusted network or device, which directly addresses preventing downloads by users outside the organization. VPC Service Controls (C) are correct because they create a service perimeter around Cloud Storage that blocks data exfiltration, including preventing data from being copied to external projects outside the perimeter. Cloud DLP (B) is incorrect because it is used to discover, classify, and redact sensitive data, not to enforce access or exfiltration boundaries.

Cloud KMS (D) is incorrect because it manages encryption keys and does not control who can download or copy data. Cloud Armor (E) is incorrect because it protects web applications from network-layer and application-layer attacks at the edge, not Cloud Storage data access or project boundaries.

492
MCQmedium

A company uses Cloud SQL for PostgreSQL and needs to run complex analytical queries on the same dataset without affecting the performance of the transactional database. What should they do?

A.Schedule periodic exports to Cloud Storage and query with BigQuery
B.Create read replicas of the Cloud SQL instance and run queries on the replicas
C.Upgrade the Cloud SQL instance to a higher machine type
D.Use BigQuery to directly query Cloud SQL via federated queries
AnswerB

Read replicas receive asynchronous copies of the primary's data, letting analytical queries run on separate compute without consuming the transactional instance's resources. This isolates workload contention, directly satisfying the requirement to avoid performance impact on the production database.

Why this answer

Option B is correct because Cloud SQL read replicas maintain a continuously updated copy of the primary instance and serve read-only traffic, so complex analytical queries can be offloaded to the replica without consuming the primary instance's CPU, memory, or I/O resources. This isolates the transactional workload from the analytical workload while keeping the same PostgreSQL dataset. Option A is not ideal because periodic exports are stale and require managing a separate BigQuery pipeline, not real-time querying of the same dataset.

Option C only adds capacity to the primary instance and does not separate analytical load from transactional traffic. Option D is unsuitable because BigQuery federated queries against Cloud SQL still send query load to the Cloud SQL instance, affecting transactional performance.

493
MCQeasy

A non-profit organization with limited IT staff wants to use cloud to improve its fundraising and donor management without hiring technology specialists. Which type of cloud service model is most appropriate for this organization's need?

A.Infrastructure as a Service (IaaS), where the organization provisions VMs and installs donor management software
B.Software as a Service (SaaS), where a fully managed donor management application is subscribed to and used without any infrastructure management
C.Platform as a Service (PaaS), where the organization deploys custom-built donor management code
D.Private cloud, where the organization builds its own cloud infrastructure for complete data control
AnswerB

SaaS is the right model. The non-profit subscribes to a ready-to-use donor management application (e.g., Salesforce Nonprofit, Bloomerang, Blackbaud) with no infrastructure to manage. Updates, security, backups, and scaling are all handled by the SaaS provider. The organization's limited IT staff can focus on using the tool, not running it.

Why this answer

Software as a Service (SaaS) is the most appropriate model because it provides a fully managed, ready-to-use donor management application over the internet. The organization's limited IT staff can simply subscribe and use the software without provisioning servers, installing applications, or managing infrastructure, directly addressing the need to avoid hiring technology specialists.

Exam trap

Google Cloud often tests the misconception that IaaS is always the 'foundation' for any cloud solution, but the trap here is that candidates overlook the organization's specific constraint of limited IT staff and choose IaaS, failing to recognize that SaaS eliminates all infrastructure and software management overhead.

How to eliminate wrong answers

Option A is wrong because IaaS requires the organization to provision and manage virtual machines, install the donor management software, and handle OS patching and scaling, which still demands significant IT expertise. Option C is wrong because PaaS requires the organization to write, deploy, and maintain custom code for the donor management application, which necessitates software development skills the organization lacks. Option D is wrong because a private cloud involves building and managing dedicated cloud infrastructure on-premises or hosted, which requires extensive IT staff for hardware, virtualization, and maintenance, contradicting the goal of avoiding technology specialists.

494
Multi-Selectmedium

A company runs a high-performance computing (HPC) workload on Compute Engine that requires low-latency, high-throughput scratch storage. The workload is checkpointed every hour. Which TWO storage options should the engineer consider for the scratch storage? (Choose 2)

Select 2 answers
A.Persistent Disk (HDD)
B.Persistent Disk (SSD)
C.Local SSD
D.Filestore
E.Cloud Storage
AnswersB, C

Persistent Disk (SSD) is a durable, network-attached block storage service that provides consistent, low-latency performance and survives instance termination. For HPC scratch, it can be used when data must persist across restarts or preemption, allowing successful checkpointing and restart without data loss. Its performance scales with provisioned size, and it supports multiple per-instance volumes to meet aggregate throughput needs.

Why this answer

Option B (Persistent Disk SSD) is correct because SSD-backed Persistent Disks deliver the high IOPS and throughput needed for low-latency scratch storage while remaining durable and independent of the VM lifecycle, which suits an HPC workload that checkpoints hourly. Option C (Local SSD) is correct because Local SSDs are physically attached NVMe devices offering the lowest latency and highest throughput per instance, making them ideal for scratch data that is regenerated from checkpoints. Option A (Persistent Disk HDD) is not appropriate because HDD-backed disks have much lower IOPS and higher latency than SSDs.

Option D (Filestore) is a managed NFS file service optimized for shared file access rather than the highest-throughput local scratch storage. Option E (Cloud Storage) is an object store accessed over the network, so it cannot provide the low-latency, high-throughput block-level scratch performance required.

Exam trap

GCDL often tests the misconception that any SSD-backed storage is equally suitable for HPC, ignoring the trade-offs between Local SSD (ephemeral, ultra-low latency) and Persistent Disk SSD (durable, slightly higher latency).

495
MCQhard

A data analytics firm wants to query data across Cloud Storage and BigQuery without moving the data. They need a single SQL interface. Which Google Cloud service enables this?

A.Dataproc with Spark SQL
B.BigQuery external tables (federated queries)
C.Cloud SQL federated queries
D.BigQuery Omni
AnswerB

BigQuery external tables, also known as federated tables, let you create a table definition that points to one or more files in Cloud Storage and query those files with standard BigQuery SQL without loading data into BigQuery storage. You can manually specify the schema or let BigQuery auto-detect it, and you can use wildcards to span multiple objects in a bucket. This directly satisfies the requirement to query data across cloud storage while keeping BigQuery as the serverless query engine.

Why this answer

BigQuery external tables (federated queries) allow SQL queries directly against data stored in Cloud Storage (CSV, JSON, Parquet, Avro, ORC) without loading it into BigQuery. This provides a single SQL interface across both native BigQuery tables and external data sources, satisfying the requirement to query without moving data.

Exam trap

GCDL often tests the difference between BigQuery external tables (GCS federation) and BigQuery Omni (AWS/Azure federation) — candidates frequently pick Omni for GCS data or assume Dataproc is needed for cross-source SQL.

How to eliminate wrong answers

Option A is wrong because Dataproc with Spark SQL requires provisioning a cluster and uses Spark SQL, not BigQuery's SQL interface, and it does not natively federate across Cloud Storage and BigQuery in a single query. Option C is wrong because Cloud SQL federated queries connect Cloud SQL to external databases, not to Cloud Storage or BigQuery. Option D is wrong because BigQuery Omni enables querying data in AWS S3 and Azure Blob Storage, not Cloud Storage, and is designed for multi-cloud rather than GCS-BigQuery federation.

496
MCQmedium

A developer needs to deploy a containerized microservice that scales to zero when not in use and automatically scales up on incoming traffic. The microservice uses a custom container image that listens on port 8080. Which Google Cloud compute service is BEST suited for this requirement?

A.Cloud Functions
B.Cloud Run
C.Google Kubernetes Engine (GKE) with Horizontal Pod Autoscaler
D.App Engine standard environment
AnswerB

Cloud Run is a fully managed serverless compute platform that executes stateless HTTP-driven containers, automatically scaling to zero when idle and scaling up instantly to handle incoming requests. It supports any container image that listens on a port, making it an ideal fit for deploying a containerized microservice without managing infrastructure, and you pay only while the container is processing requests.

Why this answer

Cloud Run is a managed compute platform that runs stateless containers, automatically scales (including to zero), and charges only for resources used during request processing. It fits the requirement perfectly. GKE requires managing a cluster and doesn't scale to zero by default; App Engine standard environment has a sandbox that may not support custom containers; Cloud Functions is limited to specific runtimes.

497
MCQmedium

A healthcare startup is building a HIPAA-compliant application on Google Cloud. They need to encrypt data at rest and manage their own encryption keys. Which service should they use for key management?

A.Cloud IAM
B.Cloud Data Loss Prevention (DLP)
C.Cloud HSM
D.Cloud Key Management Service (Cloud KMS)
AnswerD

Cloud Key Management Service (Cloud KMS) is the correct answer because it is Google Cloud's managed service for creating, rotating, and destroying customer-managed encryption keys (CMEK). The service provides a software-based key management tier suitable for HIPAA compliance, and it also integrates with Cloud HSM to offer hardware-backed keys if FIPS 140-2 Level 3 validation is required. Because you retain control over the key material lifecycle and can audit key usage, Cloud KMS meets the security and governance requirements of a HIPAA-compliant application.

Why this answer

Cloud KMS is the correct service because it is Google Cloud's dedicated key management service that allows customers to create, manage, and use encryption keys for data at rest. It supports customer-managed encryption keys (CMEK) and integrates with many GCP services, enabling HIPAA-compliant encryption where the customer controls the keys. Cloud HSM is a hardware security module option within Cloud KMS, but the core key management service is Cloud KMS.

Exam trap

The trap here is confusing Cloud HSM with Cloud KMS; candidates might think Cloud HSM is the primary key management service, but it is only a hardware-backed option within Cloud KMS.

How to eliminate wrong answers

Option A is wrong because Cloud IAM is for identity and access management, not encryption key management. Option B is wrong because Cloud DLP is used for discovering, classifying, and redacting sensitive data, not for managing encryption keys. Option C is wrong because Cloud HSM is a hardware-backed key storage option within Cloud KMS, but it is not the primary service for key management; Cloud KMS is the overarching service.

498
MCQmedium

An organization wants to use Google Cloud to analyze large-scale genomic data. The data is stored in Cloud Storage in a compressed format. They need to run a custom Python pipeline that preprocesses the data and then uses a GPU-intensive algorithm for alignment. The preprocessing is CPU-bound and takes 30 minutes per sample, while the alignment takes 1 hour per sample on a GPU. They have thousands of samples. Which compute approach is MOST cost-effective?

A.Use committed use discounts for 1 year on GPU instances
B.Use regular VMs with both CPU and GPU on the same instance
C.Use preemptible VMs with CPU for preprocessing and preemptible VMs with GPU for alignment, with checkpointing
D.Use sole-tenant nodes for data isolation
AnswerC

Preemptible VMs cost up to 70-80% less than standard VMs and are killed by Google Compute Engine within 24 hours, making them ideal for fault-tolerant batch tasks. By splitting the pipeline into CPU-based preprocessing and GPU-based alignment on separate preemptible VM groups, you avoid mixing resource types and can automatically restart interrupted tasks from checkpointed state. This minimizes compute spend while still completing the analysis, and is the most cost-effective approach for a finite workload.

Why this answer

Preemptible VMs are significantly cheaper (up to 80% discount) than regular VMs, and the workload is fault-tolerant if checkpointing is used. The preprocessing is CPU-bound and the alignment is GPU-bound, so using preemptible CPU VMs for preprocessing and preemptible GPU VMs for alignment minimizes cost. Checkpointing allows the pipeline to resume after a preemption without losing all progress, making this the most cost-effective approach for thousands of samples.

Exam trap

The trap is assuming that committed use discounts or sole-tenant nodes are cost-effective for batch workloads; candidates often overlook preemptible VMs with checkpointing as the cheapest option for fault-tolerant, interruptible jobs.

How to eliminate wrong answers

Option A is wrong because committed use discounts require a 1-year commitment and still cost more than preemptible VMs for a batch workload that can tolerate interruptions; they are better for steady-state, long-running workloads. Option B is wrong because using regular VMs with both CPU and GPU on the same instance is more expensive than using separate preemptible instances tailored to each phase, and it does not leverage the cost savings of preemptible VMs. Option D is wrong because sole-tenant nodes are for physical isolation and compliance, not cost optimization; they are significantly more expensive and do not address the cost-effectiveness requirement.

499
MCQmedium

A company is migrating a legacy monolithic application to Google Cloud. The application currently runs on a single physical server with a custom Linux distribution. The team wants to minimize changes to the application while gaining the benefits of cloud infrastructure. Which migration strategy should they use?

A.Rehost the application on Compute Engine using a custom image of the current OS
B.Rebuild the application from scratch on Cloud Run
C.Re-platform the application to a supported OS version
D.Refactor the application into microservices and deploy on GKE
AnswerA

Rehosting on Compute Engine with a custom image of the current OS is a classic 'lift-and-shift' strategy that preserves the exact kernel, libraries, and configuration of the legacy environment. Because no code changes or architectural changes are required, this approach minimizes migration risk and time-to-completion, allowing the team to move the application as-is. The custom image ensures parity with the on-premises OS, avoiding issues with unsupported or altered system dependencies.

Why this answer

Lift and shift (rehosting) moves the application as-is to the cloud, often by creating a custom image to run on Compute Engine. This minimizes changes. Refactoring (rearchitecting) involves code changes; re-platforming modifies the OS/platform; rebuilding is a full rewrite.

500
MCQmedium

A company runs batch processing jobs on scheduled intervals. They want to minimise costs by using short-lived compute capacity that can be interrupted but offers significant discounts. Which type of Compute Engine VM should they use?

A.E2 high-memory VMs
B.Sole-tenant nodes
C.Preemptible VMs
D.Confidential VMs
AnswerC

Preemptible VMs (and Spot VMs) are Compute Engine instances that can be terminated anywhere anytime due to excess capacity reuse, but they can be used for batch and fault-tolerant workloads. They are up to 60–80% cheaper than standard VMs, making them the ideal choice for a company running scheduled batch processing that can checkpoint and resume. To use them effectively, the application must handle unexpected termination gracefully and be restartable from saved state.

Why this answer

Preemptible VMs are short-lived Compute Engine instances that offer up to 60-91% discounts compared to standard VMs but can be terminated by Google at any time with a 30-second notice. They are ideal for batch processing jobs that are fault-tolerant and can be interrupted, matching the requirement to minimize costs with interruptible capacity.

Exam trap

GCDL often tests the distinction between preemptible/spot VMs and other VM types — candidates may pick high-memory or confidential VMs based on workload characteristics, missing that the key requirement is interruptible, discounted capacity.

How to eliminate wrong answers

Option A is wrong because E2 high-memory VMs are standard on-demand instances optimized for memory-intensive workloads, not discounted interruptible capacity. Option B is wrong because sole-tenant nodes provide physical isolation for compliance or licensing needs and are more expensive, not a cost-saving interruptible option. Option D is wrong because Confidential VMs encrypt data in use with AMD SEV and are priced at a premium for security, not for cost savings on batch jobs.

501
MCQeasy

Which defense-in-depth layer includes measures like access controls, vulnerability management, and intrusion detection systems?

A.Data security
B.Physical security
C.Operational security
D.Infrastructure security
AnswerC

Operational security is correct because it encompasses the day-to-day processes and controls that protect systems, including user access management (authentication, authorization, least privilege), vulnerability management, continuous monitoring, and incident response. These are precisely the measures described in the question—access control lists and intrusion detection—which require ongoing operational discipline rather than static architectural safeguards. This layer ensures that policies and procedures are executed consistently across the environment.

Why this answer

Operational security (OpSec) is the defense-in-depth layer focused on day-to-day processes, people, and procedures that protect systems — it explicitly encompasses access controls, vulnerability management, and intrusion detection systems. These are ongoing operational activities rather than physical barriers or data-at-rest protections, which is why OpSec is the correct classification.

Exam trap

The trap here is conflating infrastructure security with operational security — candidates see 'intrusion detection' and pick infrastructure, but the exam expects you to recognize that access control processes, vulnerability management, and IDS monitoring are operational activities, not just infrastructure hardening.

How to eliminate wrong answers

Option A is wrong because data security focuses on protecting data at rest and in transit (encryption, tokenization, DLP, key management), not on IDS or vulnerability scanning workflows. Option B is wrong because physical security covers tangible controls like fences, locks, guards, CCTV, and badge readers — it does not include vulnerability management or IDS. Option D is wrong because infrastructure security refers to hardening the underlying network, hosts, and cloud infrastructure (firewalls, segmentation, patching baselines), which is a component of but not synonymous with the broader operational security layer that also covers process and personnel controls.

502
MCQmedium

A company's finance team wants to understand why their cloud bills vary significantly month to month, unlike their fixed on-premises IT costs. Which fundamental cloud pricing characteristic explains this variability?

A.Cloud providers change their prices frequently, causing unpredictable costs
B.Consumption-based pricing means cloud costs scale directly with actual usage, unlike fixed on-premises costs
C.Cloud providers apply hidden fees that vary randomly each month
D.Cloud costs are fixed like on-premises costs; the variability must be caused by billing errors
AnswerB

This is the correct explanation. Cloud is utility-like pricing: a compute-heavy month costs more than a quiet month. Finance teams must shift from thinking about fixed IT budgets to variable cost management tied to business activity levels.

Why this answer

Cloud computing operates on a consumption-based (pay-as-you-go) pricing model, where costs are directly tied to the amount of resources consumed (e.g., compute hours, storage GB, data transfer). Unlike fixed on-premises IT costs, which are incurred regardless of actual usage (e.g., hardware depreciation, facility leases), cloud bills fluctuate as usage scales up or down. This fundamental characteristic explains the month-to-month variability observed by the finance team.

Exam trap

Google Cloud often tests the misconception that cloud pricing is unpredictable or error-prone, when in fact the variability is a deliberate feature of consumption-based pricing, not a flaw or hidden fee.

How to eliminate wrong answers

Option A is wrong because cloud providers do not change their prices frequently; instead, they typically announce price reductions or new tiers well in advance, and pricing is stable over short periods. Option C is wrong because cloud providers are transparent about their pricing models and do not apply hidden fees that vary randomly; all charges are itemized in the billing dashboard based on metered usage. Option D is wrong because cloud costs are not fixed like on-premises costs; the variability is a direct result of consumption-based pricing, not billing errors, and cloud billing systems are highly accurate.

503
MCQhard

A data team has an IAM policy on a BigQuery dataset as shown. Alice needs to run a query that joins across multiple datasets. She receives a permission error. What is the most likely cause?

A.The policy denies all users except Bob
B.Alice lacks the jobUser role to run queries
C.Alice does not have permission to read the dataset
D.Bob’s dataOwner role prevents others from querying
AnswerB

Running a query in BigQuery requires two permission sets: data access, such as roles/bigquery.dataViewer, and job execution permission, such as roles/bigquery.jobUser, to create a query job. Alice has dataViewer on the dataset, but that role lacks the bigquery.jobs.create permission at the project or job level. Without jobUser (or a broader role like roles/bigquery.user), her query requests fail even though she can read table data. This is the specific reason her query fails.

Why this answer

The BigQuery `jobUser` role is required to run query jobs, including those that join across datasets. The IAM policy shown only grants dataset-level permissions (like `dataViewer` or `dataOwner`), but Alice lacks the `jobUser` role at the project level, which is necessary to submit a query job. Without this role, she receives a permission error even if she has read access to the datasets.

Exam trap

The trap here is that candidates assume dataset-level read permissions (like `dataViewer`) are sufficient to run queries, but BigQuery requires the separate `jobUser` role at the project level to execute query jobs.

How to eliminate wrong answers

Option A is wrong because the policy does not deny all users except Bob; it only grants specific roles to Bob and others, and does not include an explicit deny statement. Option C is wrong because the error occurs when joining across datasets, which requires the `jobUser` role to run the query job, not just read permission on the dataset. Option D is wrong because Bob's `dataOwner` role does not prevent others from querying; it grants full control over the dataset but does not block other users' permissions.

504
MCQhard

A company runs a batch processing workload every night that takes 60 minutes on a single n1-standard-32 VM. They want to reduce costs by using preemptible VMs but need the job to complete within 90 minutes. The job can be parallelized if necessary. Which approach is most cost-effective?

A.Use a single non-preemptible VM to ensure completion.
B.Use the same VM but as preemptible; if preempted, restart the job from scratch.
C.Split the workload across 4 preemptible n1-standard-8 VMs with checkpointing.
D.Use a single preemptible VM with persistent disk snapshots every 10 minutes.
AnswerC

Splitting the workload across 4 preemptible n1-standard-8 VMs reduces per-VM runtime to roughly one-quarter of the original, assuming the batch is parallelizable, which helps fit within the 90-minute window. Each VM writes checkpoints of its progress to a shared persistent location (e.g., Cloud Storage or a persistent disk); if any VM is preempted, a replacement VM can be launched and resume from the last saved state, losing only minutes of work rather than the entire job. Preemptible VMs cost about 20-40% of standard VMs, so the solution is highly cost-effective while still being resilient to preemption via checkpointing.

Why this answer

Preemptible VMs can be terminated at any time, so running multiple smaller instances with checkpointing and retries reduces cost and meets the deadline.

505
MCQmedium

A company uses multiple public clouds (AWS, Azure, Google Cloud) for different workloads. They want to centralize monitoring and logging. Which Google Cloud service can aggregate logs from all clouds?

A.Cloud Audit Logs
B.Cloud Console
C.Cloud Monitoring
D.Cloud Logging
AnswerD

Cloud Logging is the correct answer because it is a fully managed, real-time log aggregation service capable of ingesting logs from any source, including AWS and Azure, via the Logging API, Fluent Bit agents, or SDK integrations. It accepts structured and unstructured logs, applies indexes for queryability, and supports exports to BigQuery, Pub/Sub, or cloud storage for long-term archiving. In a multi-cloud environment, you can install an agent on AWS EC2 or Azure VMs to forward logs to a centralized sink, giving you a single Logs Explorer to search across all clouds. This directly fulfills the requirement of aggregating logs from multiple public clouds.

Why this answer

Cloud Logging (formerly Stackdriver Logging) is Google Cloud's centralized log ingestion, storage, and analysis service, and it can ingest logs from AWS, Azure, and other sources via the BindPlane/Cloud Logging agent or the Logging API. It provides a single pane for querying and routing logs across clouds, making it the correct choice for aggregating multi-cloud logs. Cloud Monitoring handles metrics and dashboards, not log aggregation.

Exam trap

GCDL often tests the boundary between Cloud Logging (logs) and Cloud Monitoring (metrics), and whether candidates know Cloud Audit Logs are Google-Cloud-only and cannot aggregate AWS/Azure logs.

How to eliminate wrong answers

Option A is wrong because Cloud Audit Logs capture administrative and data access activity within Google Cloud only — they do not aggregate logs from AWS or Azure. Option B is wrong because Cloud Console is a web UI for managing Google Cloud resources, not a log aggregation service. Option C is wrong because Cloud Monitoring focuses on metrics, uptime checks, dashboards, and alerting; while it integrates with Cloud Logging, it is not the log aggregation service itself.

506
MCQmedium

A company needs to set a budget for their GCP project and receive notifications when spending reaches 50%, 90%, and 100% of the budget. Which action should they take?

A.Enable Committed Use Discounts to reduce costs.
B.Set up billing export to BigQuery with scheduled queries.
C.Configure a budget alert in Cloud Billing with threshold rules at 50%, 90%, and 100%.
D.Use Active Assist to set cost thresholds.
AnswerC

Configuring a budget alert in Cloud Billing is the purpose-built and recommended way to receive spending alerts. You can define a budget amount (for example, a monthly cap) and set threshold rules at 50%, 90%, and 100% to trigger notifications via email or Pub/Sub when actual or forecasted costs hit those percentages. This causes the Cloud Billing system to automatically send alerts, addressing the company's need to be notified as costs approach the budget. It is the only option that explicitly provides the required alerting behavior.

Why this answer

The correct action is to configure a budget alert in Cloud Billing with threshold rules at 50%, 90%, and 100%. Budget alerts in GCP allow you to set multiple thresholds on a budget, and when spending reaches each threshold, notifications are sent via email or Pub/Sub. This directly meets the requirement of receiving notifications at specific percentages of the budget.

Exam trap

The trap here is confusing cost optimization tools (like Committed Use Discounts) or analysis tools (like BigQuery export) with budget alerting, when the question specifically asks for notifications at spending thresholds.

How to eliminate wrong answers

Option A is wrong because Committed Use Discounts (CUDs) are a pricing mechanism to reduce costs by committing to a certain level of usage, not a notification system for budget thresholds. Option B is wrong because billing export to BigQuery with scheduled queries is a method for detailed cost analysis and custom reporting, but it does not natively provide real-time notifications at specific budget thresholds without additional setup and code. Option D is wrong because Active Assist is a suite of tools that provides recommendations and insights for optimizing GCP resources, not a budgeting or alerting service for cost thresholds.

507
MCQeasy

A developer wants to deploy a containerized application that can scale down to zero when not in use and charges only for the resources consumed during request processing. Which Google Cloud compute service should they choose?

A.Google Kubernetes Engine (GKE)
B.Compute Engine
C.App Engine Flexible Environment
D.Cloud Run
AnswerD

Cloud Run is a fully managed serverless container platform that automatically scales your container from zero instances up to a number sufficient to handle incoming HTTP requests, and then back down to zero when traffic disappears. You are billed only for the exact time spent processing a request, with a minimum granularity of 100 milliseconds, so when there is no request traffic there are zero compute charges. This gives the developer exactly what they asked for: a containerized application that scales to zero, eliminating infrastructure costs during idle periods.

Why this answer

Cloud Run is a fully managed serverless container platform that automatically scales instances down to zero when there is no traffic, and it bills only for CPU/memory consumed during request handling (with per-100ms granularity). This matches the requirement of zero-idle cost and pay-per-request processing exactly. GKE and Compute Engine both require at least one running node/VM, and App Engine Flexible keeps at least one instance running, so none of them can truly scale to zero.

Exam trap

GCDL often tests the distinction between 'serverless' services that still keep a warm instance (App Engine Flexible) versus those that truly scale to zero (Cloud Run, Cloud Functions) — candidates who equate 'serverless' with 'scale-to-zero' pick App Engine Flexible incorrectly.

How to eliminate wrong answers

Option A is wrong because GKE clusters require a node pool with at least one running node, so the cluster never scales to zero and you pay for idle node capacity. Option B is wrong because Compute Engine VMs are always-on resources billed per second while running; they cannot scale to zero without external orchestration and still incur persistent disk costs. Option C is wrong because App Engine Flexible Environment keeps a minimum of one instance running at all times and does not support true scale-to-zero billing.

508
MCQmedium

A media company has a web application that serves video content globally. The application is deployed on Compute Engine instances behind a TCP load balancer in a single region. Users in distant regions experience high latency. The company wants to improve performance for all users while keeping operational overhead low. They also need to handle sudden spikes in traffic during live events. What should they do?

A.Deploy additional instances in multiple regions and use a global HTTP(S) load balancer with Cloud CDN.
B.Use Cloud Run for the application and enable automatic scaling globally.
C.Move the application to Google Kubernetes Engine and use horizontal pod autoscaling.
D.Increase the machine type of existing instances and add more instances in the same region.
AnswerA

Deploying additional regional instances behind a global HTTP(S) load balancer uses Google's anycast IP to route each user to the nearest healthy backend, dramatically cutting network round-trip time. Cloud CDN then caches video segments at edge locations, so repeated views are served from the edge, not the origin, reducing latency and origin load. This combination uniquely provides both global latency reduction and the ability to absorb traffic spikes by distributing capacity across multiple regions.

Why this answer

Deploying instances in multiple regions and using a global HTTP(S) load balancer with Cloud CDN reduces latency by serving content from edge locations close to users. Cloud CDN caches video content at Google's global edge points of presence (PoPs), while the global HTTP(S) load balancer provides anycast IP-based traffic distribution across regions, handling traffic spikes through automatic scaling and distributed capacity.

Exam trap

The trap here is that candidates may think Cloud Run or GKE with autoscaling alone can solve global latency, but they overlook the need for multi-region deployment and edge caching, which are essential for reducing geographic latency and handling global traffic spikes with low operational overhead.

How to eliminate wrong answers

Option B is wrong because Cloud Run does not support automatic scaling globally across multiple regions; it is a regional service and would require manual multi-region setup or additional services like a multi-cluster ingress, increasing operational overhead. Option C is wrong because moving to Google Kubernetes Engine with horizontal pod autoscaling only addresses scaling within a single cluster and does not solve global latency or provide multi-region load balancing without additional complex configuration. Option D is wrong because increasing machine types and adding instances in the same region does not reduce latency for distant users; it only improves capacity within that single region, failing to address geographic distance.

509
MCQeasy

What is the primary purpose of VPC Service Controls?

A.To control ingress and egress traffic at the network level
B.To detect network intrusions
C.To prevent data exfiltration from Google Cloud services
D.To protect against DDoS attacks
AnswerC

VPC Service Controls create perimeters around Google Cloud APIs and services, using context such as identity, source IP, and resource metadata to authorize access. This prevents copying or moving data from within the perimeter to external destinations, directly reducing the risk of data theft. By working at the API layer, it complements IAM and network controls with a data-boundary layer.

Why this answer

VPC Service Controls create perimeters around Google Cloud services to prevent data exfiltration. VPC firewall rules control network traffic. Cloud Armor is for DDoS.

Cloud IDS is for intrusion detection.

510
MCQmedium

A company's PostgreSQL database has grown to 50 TB and their application requires near-zero downtime, automatic failover, and the ability to scale reads horizontally without the migration complexity of switching to Spanner. Which Google Cloud database product is specifically designed as a fully managed, highly scalable PostgreSQL-compatible database?

A.Cloud SQL (PostgreSQL)
B.AlloyDB for PostgreSQL
C.Cloud Spanner
D.Bare metal PostgreSQL on Compute Engine
AnswerB

AlloyDB for PostgreSQL is the correct choice because it is a fully managed, PostgreSQL-compatible database engineered for demanding enterprise workloads. It delivers up to 4× faster online transaction processing (OLTP) and up to 100× faster analytical queries than standard PostgreSQL, while providing 99.99% availability and horizontal read scaling via built-in columnar engine and read pools. Because it is wire-compatible with PostgreSQL, existing application code, drivers, and tools work unchanged — making it the only option that combines full PostgreSQL compatibility with the scale and performance needed for 50 TB of data.

Why this answer

AlloyDB for PostgreSQL is a fully managed, PostgreSQL-compatible database service designed for high scalability, near-zero downtime, and automatic failover. It separates compute and storage to enable horizontal read scaling with read pools, and it uses a columnar engine for analytical acceleration, making it ideal for large workloads like 50 TB without the migration complexity of Spanner.

Exam trap

The trap here is that candidates confuse Cloud SQL's PostgreSQL offering with AlloyDB's PostgreSQL compatibility, overlooking Cloud SQL's storage and scaling limitations for large, high-availability workloads.

How to eliminate wrong answers

Option A is wrong because Cloud SQL for PostgreSQL is limited to 30 TB of storage and does not support automatic horizontal read scaling or near-zero downtime failover at the scale of 50 TB. Option C is wrong because Cloud Spanner is a globally distributed, strongly consistent database that is not PostgreSQL-compatible and requires significant application migration to change from PostgreSQL semantics. Option D is wrong because bare metal PostgreSQL on Compute Engine is not a fully managed service; it requires manual configuration for failover, scaling, and maintenance, and does not provide the automatic, near-zero downtime capabilities specified.

511
Drag & Dropmedium

Drag and drop the steps to recover a Compute Engine VM from a snapshot in the correct order.

Drag or tap steps into the slots.

Steps
Order
1Step 1
2Step 2
3Step 3
4Step 4

Why this order

The recovery process involves using the snapshot to create a disk, detaching the old boot disk, attaching the new one, and starting the VM.

512
MCQhard

An organization uses labels to track cost by environment (dev, test, prod). However, the Finance team notices that some resources are missing the 'env' label. Which service can automatically suggest labeling resources that are missing required labels?

A.Cost Management dashboard
B.Active Assist Recommendations
C.Cloud Asset Inventory
D.Organization Policy Service
AnswerB

Active Assist Recommendations includes the label recommender, which leverages machine learning to analyze resource metadata, usage patterns, and existing label coverage to identify resources missing required label keys or values. It surfaces each suggestion with a rationale and one-click remediation, helping teams enforce tagging conventions without manual audits. This is exactly the kind of proactive, prescriptive guidance the organization needs.

Why this answer

Active Assist provides recommendations for various optimizations, including labeling. The Recommender can suggest labels for resources that are missing them, based on usage patterns.

513
MCQhard

A security administrator needs to ensure that Google personnel do not access customer data without explicit authorization. Which service should they use to get logs of Google employee access?

A.Access Transparency
B.Cloud Audit Logs
C.Security Command Center
D.Assured Workloads
AnswerA

Access Transparency is the correct control because it provides near real-time, read-only audit logs specifically for actions taken by Google personnel against customer data. These logs capture each instance of access (e.g., when a Google engineer troubleshoots a support issue) and are viewable in Cloud Logging. Unlike standard audit logs, Access Transparency is designed solely for visibility into Google's own staff activity, making it the exact mechanism for this requirement.

Why this answer

Access Transparency provides logs of Google personnel actions on customer data. Cloud Audit Logs track user activities within the customer's project. Assured Workloads is for regulatory compliance.

Security Command Center is for threat detection.

514
MCQhard

A machine learning team wants to train, evaluate, deploy, and monitor ML models in a unified platform without managing infrastructure, and with built-in support for experiment tracking, model versioning, and A/B testing between model versions. Which Google Cloud product provides this end-to-end managed ML platform?

A.BigQuery ML, for training and deploying ML models using SQL within BigQuery
B.Vertex AI, Google Cloud's unified ML platform covering training, experiment tracking, model registry, deployment, and monitoring in a single managed service
C.Cloud Dataproc, for running distributed Spark ML jobs on managed Hadoop clusters
D.Cloud Functions, for deploying ML inference code as serverless functions
AnswerB

Vertex AI is the complete answer. It provides: managed training (custom containers or AutoML), Vertex AI Experiments (experiment tracking and comparison), Vertex AI Model Registry (version management), Vertex AI Endpoints (serving with traffic splitting for A/B testing), and Model Monitoring (data drift and skew detection). This is Google Cloud's end-to-end ML platform.

Why this answer

Vertex AI is Google Cloud's unified ML platform that provides an end-to-end managed service for training, evaluating, deploying, and monitoring ML models without requiring infrastructure management. It includes built-in experiment tracking, a model registry for versioning, and supports A/B testing between model versions, directly matching the question's requirements.

Exam trap

The trap here is that candidates may confuse BigQuery ML's SQL-based model training with a full ML platform, overlooking its lack of experiment tracking, model versioning, and A/B testing capabilities that Vertex AI provides.

How to eliminate wrong answers

Option A is wrong because BigQuery ML is limited to training and deploying models using SQL within BigQuery, lacking built-in experiment tracking, model versioning, and A/B testing capabilities for custom ML workflows. Option C is wrong because Cloud Dataproc is a managed Spark and Hadoop service for distributed data processing, not a unified ML platform with experiment tracking, model registry, or A/B testing features. Option D is wrong because Cloud Functions is a serverless compute service for event-driven code execution, not designed for ML model training, experiment tracking, or A/B testing between model versions.

515
MCQhard

An organization uses Security Command Center (SCC) premium tier and wants to automatically remediate a specific finding type by disabling public access to Cloud Storage buckets. What is the recommended approach?

A.Use Cloud Pub/Sub alone to listen for SCC findings and then manually remediate
B.Use IAM to deny all users except project owners from making buckets public
C.Set up a Cloud Function triggered by SCC findings to remove public access
D.Configure a Cloud Scheduler job to run a script that checks and removes public access
AnswerC

This is the correct event-driven pattern: SCC publishes findings to Cloud Pub/Sub via a notification config, and a Cloud Function can be subscribed to that Pub/Sub topic using the background-trigger mechanism. The function receives the finding JSON, parses the bucket name, and uses the Cloud Resource Manager or Cloud Asset API to remove the public IAM binding or the public ACL. This happens within milliseconds of detection, is fully automated, and scales naturally with findings, making it the right fit for real-time security remediation.

Why this answer

Security Command Center (SCC) premium tier can publish findings to Cloud Pub/Sub, which can trigger a Cloud Function via a push subscription. The Cloud Function can then use the Google Cloud Storage API (specifically, the `storage.buckets.setIamPolicy` method) to remove the `allUsers` or `allAuthenticatedUsers` bindings that grant public access, enabling automated, event-driven remediation without manual intervention.

Exam trap

Google Cloud often tests the distinction between event-driven automation (Cloud Functions + Pub/Sub) and scheduled or manual approaches, so candidates mistakenly choose Cloud Scheduler or IAM deny policies because they think 'automation' means periodic checks or preventive controls, rather than reactive, real-time remediation.

How to eliminate wrong answers

Option A is wrong because using Cloud Pub/Sub alone only delivers the finding notification; it does not perform any remediation action, so manual steps are still required, which defeats the goal of automation. Option B is wrong because IAM deny policies cannot retroactively remove public access from buckets that already have public bindings; they only prevent future grants, and they do not automatically remediate existing public buckets. Option D is wrong because Cloud Scheduler runs on a fixed schedule, not in response to SCC findings, so it introduces latency and cannot provide real-time remediation when a finding is generated.

516
Matchingmedium

Match each Google Cloud service to its primary use case.

Drag a concept onto its matching description — or click a concept then click the description.

Concepts
Matches

Virtual machines (IaaS)

Unstructured object storage

Serverless data warehouse

Serverless container platform

Event-driven serverless functions

Why these pairings

Compute Engine provides IaaS virtual machines, Cloud Storage offers object storage for unstructured data, and BigQuery is a data warehouse for analytics. Common confusions occur when mixing these services.

517
MCQmedium

A company is planning a cloud migration and wants to understand the difference between 'lift and shift' and 'cloud-native' approaches. Which statement correctly distinguishes these two migration strategies?

A.Lift and shift uses containers; cloud-native uses virtual machines.
B.Lift and shift moves applications to the cloud with minimal changes; cloud-native re-architects applications to leverage cloud-specific features and managed services.
C.Lift and shift is only possible for new applications; cloud-native is for existing applications.
D.Lift and shift costs more long-term; cloud-native costs more short-term due to licensing.
AnswerB

Lift-and-shift migrates applications to the cloud with minimal or no changes, often copying VMs or servers as-is to IaaS, which is fast but fails to fully utilize cloud capabilities like autoscaling, managed databases, or serverless. Cloud-native, in contrast, re-architects applications using microservices, containers, and managed services to maximize scalability, resiliency, and operational efficiency. This is the fundamental strategic distinction: minimal relocation versus intentional redesign to leverage cloud-specific features.

Why this answer

'lift and shift' (rehosting) involves moving applications to the cloud with minimal or no changes, often using Infrastructure as a Service (IaaS) to replicate the on-premises environment. In contrast, 'cloud-native' (re-architecting) redesigns applications to use cloud-specific features like auto-scaling, managed databases, and serverless compute, fully leveraging the cloud's elasticity and pay-as-you-go model.

Exam trap

The trap here is that candidates often confuse the tools (containers vs. VMs) with the strategy, assuming lift and shift always uses containers and cloud-native uses VMs, when in fact the opposite is true for typical implementations.

How to eliminate wrong answers

Option A is wrong because lift and shift typically uses virtual machines (VMs) to replicate on-premises infrastructure, not containers; cloud-native applications often use containers (e.g., Docker) and orchestration (e.g., Kubernetes) for microservices, not VMs. Option C is wrong because lift and shift is primarily used for existing legacy applications to migrate quickly, while cloud-native is often applied to new applications but can also involve re-architecting existing ones. Option D is wrong because lift and shift can lead to higher long-term costs due to inefficient resource utilization and licensing, while cloud-native may have higher initial development costs but lower operational costs over time due to optimized resource usage and managed services.

518
MCQmedium

A company uses Google Workspace for identity. They want employees to use their Google Workspace credentials to access third-party applications (Salesforce, Slack, etc.) without separate passwords for each app. Which technology enables this?

A.VPN — employees connect to the corporate VPN which provides access to all apps.
B.Single Sign-On (SSO) using SAML 2.0 or OIDC with Google Workspace as the Identity Provider.
C.Cloud Armor — blocks unauthorized access attempts to applications.
D.Shared service account — all employees use the same credential.
AnswerB

Google Workspace acts as the identity provider (IdP), authenticating the employee once and then passing a SAML 2.0 assertion or OIDC ID token to each SaaS application, which acts as a service provider. Because the assertion/token is cryptographically signed and trusted by the application, the user is granted access without needing to re-enter credentials or create a separate password. This is the correct approach because it decouples authentication from application access, enabling centralized lifecycle management and enforcing corporate MFA policies across all federated apps.

Why this answer

Single Sign-On (SSO) using SAML 2.0 or OIDC allows Google Workspace to act as the Identity Provider (IdP), issuing authentication tokens that third-party applications (like Salesforce and Slack) trust. This eliminates the need for separate passwords, as users authenticate once with Google Workspace and the IdP handles subsequent access via security assertions or ID tokens.

Exam trap

Google Cloud often tests the distinction between network-level access (VPN) and identity-level federation (SSO), so candidates mistakenly choose VPN because they think it 'provides access to all apps' without realizing it does not solve the separate-password problem.

How to eliminate wrong answers

Option A is wrong because a VPN provides network-layer access to corporate resources but does not eliminate separate application passwords; users still need to authenticate to each app individually. Option C is wrong because Cloud Armor is a web application firewall and DDoS protection service that blocks malicious traffic at the edge, not an identity federation or SSO technology. Option D is wrong because a shared service account violates the principle of least privilege and non-repudiation; all employees using the same credential would create severe security and audit issues, and it does not enable passwordless access to third-party apps.

519
MCQeasy

A manufacturing company wants to use sensor data from equipment to predict failures before they happen, reducing downtime. How does cloud technology enable this transformation?

A.By using edge computing to store all data locally
B.By setting up monitoring alerts for equipment failure
C.By migrating all data to a data warehouse for batch analysis
D.By offering IoT Core and AI Platform to collect data and build predictive models
AnswerD

IoT Core ingests telemetry from equipment sensors, while AI Platform trains predictive models on that data to forecast failures before downtime occurs. This directly satisfies the stem's requirement for predictive maintenance, since the cloud supplies both managed ingestion at scale and the machine-learning tooling needed to turn raw sensor readings into actionable failure predictions.

Why this answer

Option D is correct because cloud providers offer managed IoT services such as AWS IoT Core or Azure IoT Hub to ingest sensor telemetry at scale, paired with AI/ML platforms (e.g., Amazon SageMaker, Azure Machine Learning) to train predictive maintenance models that forecast failures before they occur. This combination directly enables the transformation from reactive to predictive maintenance, reducing unplanned downtime. Option A is wrong because storing all data locally with edge computing limits centralized analytics and model training, and edge computing is typically used for low-latency filtering, not bulk storage.

Option B is insufficient because monitoring alerts are reactive threshold notifications, not predictive modeling. Option C is wrong because batch analysis in a data warehouse does not provide the real-time streaming ingestion and ML model deployment needed for failure prediction.

520
MCQmedium

A company wants to migrate its existing on-premises virtual machines (VMware VMs) to Google Cloud with minimal changes to the operating system and applications. Which Google Cloud product is specifically designed for migrating on-premises VMs to Google Cloud with minimal modification?

A.Migrate to Virtual Machines (formerly Velostrata), which migrates on-premises VMware VMs to Compute Engine with minimal modification and downtime
B.Cloud Dataflow, by streaming data from on-premises VMs to Google Cloud storage
C.Anthos, by registering on-premises Kubernetes clusters with Google Cloud's management plane
D.Cloud Storage Transfer Service, by copying VM disk images from on-premises storage to Cloud Storage
AnswerA

Migrate to Virtual Machines is the purpose-built service for this. It performs VM migrations from VMware (and other sources) to Compute Engine, handling the OS and application translation automatically. The 'minimal changes' requirement is the defining characteristic — it's a lift-and-shift migration tool.

Why this answer

Migrate to Virtual Machines (formerly Velostrata) is the correct choice because it is specifically designed to migrate on-premises VMware VMs to Compute Engine with minimal modification to the OS and applications. It uses a streaming migration approach that moves the VM's disk state incrementally while keeping the VM running, resulting in minimal downtime and no need to re-architect the workloads.

Exam trap

Google Cloud often tests the distinction between general-purpose data transfer or processing services (like Dataflow or Storage Transfer Service) and specialized migration tools, tempting candidates to pick a familiar service that sounds plausible but lacks the specific VM migration capability.

How to eliminate wrong answers

Option B is wrong because Cloud Dataflow is a stream and batch data processing service, not a VM migration tool; it cannot migrate entire VMs with their OS and applications intact. Option C is wrong because Anthos focuses on managing Kubernetes clusters across environments, not on migrating individual VMware VMs with minimal modification. Option D is wrong because Cloud Storage Transfer Service is designed for bulk data transfers to Cloud Storage, not for live VM migration; it would require manual disk image creation and does not handle OS/application state or minimize downtime.

521
MCQmedium

A startup needs to quickly deploy a web application with minimal infrastructure management. They want to focus on code, not servers. Which Google Cloud service model is most appropriate?

A.Software as a Service (SaaS) using Gmail
B.Function as a Service (FaaS) using Cloud Functions
C.Infrastructure as a Service (IaaS) using Compute Engine
D.Platform as a Service (PaaS) using App Engine
AnswerD

App Engine is a PaaS that abstracts away the underlying servers, OS, and networking; you simply deploy your code and it automatically handles scaling, load balancing, health checks, and runtime patching. It supports standard frameworks and routes HTTP traffic to your application without any infrastructure provisioning. This matches the need for a quick deployment while keeping your team focused on application code rather than operations.

Why this answer

App Engine is a fully managed Platform as a Service (PaaS) that abstracts away the underlying infrastructure, allowing developers to deploy web applications without managing servers or operating systems. It automatically handles scaling, load balancing, and patching, which aligns with the startup's requirement to focus on code rather than infrastructure management.

Exam trap

Google Cloud often tests the distinction between PaaS and FaaS by presenting a scenario that requires a full web application, where candidates mistakenly choose FaaS (Cloud Functions) because they confuse 'serverless' with 'no infrastructure management,' ignoring that FaaS is unsuitable for long-running HTTP applications with stateful sessions.

How to eliminate wrong answers

Option A is wrong because Gmail is a Software as a Service (SaaS) application, not a service model for deploying custom web applications; it provides no platform for code deployment or customization. Option B is wrong because Cloud Functions is a Function as a Service (FaaS) designed for event-driven, stateless functions, not for hosting a complete web application with persistent HTTP routing and session management. Option C is wrong because Compute Engine is Infrastructure as a Service (IaaS) that requires manual provisioning, configuration, and management of virtual machines, which contradicts the goal of minimizing infrastructure management.

522
MCQmedium

An engineer needs to analyze detailed cost data, including resource-level usage costs and labels, to create custom reports. Which approach should they use?

A.Use the Cost Management dashboard to export a CSV
B.Set up billing export to BigQuery
C.Manually review invoices
D.Use the Cloud Pricing Calculator
AnswerB

Setting up billing export to BigQuery is the correct approach because it continuously streams all billing data, including line items with resource IDs, labels, and usage metrics, into a BigQuery dataset. The exported tables offer full SQL queryability, enabling you to join cost data with other enterprise datasets, build custom dashboards, and perform granular analysis such as grouping by label or cost per resource. This is the only method that provides the flexibility and detail required for the engineer's needs.

Why this answer

Setting up billing export to BigQuery provides detailed cost data, including resource-level usage costs and labels, enabling custom analysis and reporting. BigQuery allows you to query and join billing data with other datasets, and it includes labels for cost allocation.

Exam trap

GCDL often tests the difference between the Cost Management dashboard (aggregated) and billing export to BigQuery (detailed, resource-level).

How to eliminate wrong answers

Option A is wrong because the Cost Management dashboard provides aggregated views and CSV export, but it does not include the granular resource-level data and labels needed for custom reports. Option C is wrong because manually reviewing invoices is not scalable and lacks detailed resource-level data. Option D is wrong because the Cloud Pricing Calculator estimates costs for planned usage, not actual detailed cost data.

523
MCQeasy

Refer to the exhibit. A security administrator reviews this Cloud Audit Logs entry. What does this entry indicate?

A.The user attempted to read the object 'secret.pdf' and the request resulted in an error.
B.The user updated the IAM policy on the bucket.
C.The user attempted to delete the object 'secret.pdf'.
D.The user successfully read the object 'secret.pdf'.
AnswerA

The audit entry records a data-access operation on the object secret.pdf, with a status field showing the request failed. This confirms the user attempted a read and the operation returned an error rather than succeeding.

Why this answer

The correct option is A: the user attempted to read the object 'secret.pdf' and the request resulted in an error. In Cloud Audit Logs, a data-access entry for a storage object typically records a method such as storage.objects.get, which corresponds to reading/downloading an object, and the presence of an error/status field (for example, a non-OK status or error code) indicates the request failed rather than succeeded. Option D is wrong because a successful read would show a successful status without an error result.

Option B is wrong because updating an IAM policy would appear as a setIamPolicy operation on the bucket, not an object read. Option C is wrong because deleting an object would be logged as storage.objects.delete, not a read/get operation.

524
MCQhard

A global gaming company uses Cloud Spanner for their leaderboard. They notice that write latency spikes during peak hours. The database is currently deployed in a single region. Which scaling strategy should they implement to reduce write latency globally?

A.Use Cloud Spanner multi-region configuration.
B.Implement application-level caching with Memorystore.
C.Change to Cloud Bigtable for higher throughput.
D.Add more nodes to the existing Spanner instance.
AnswerA

Cloud Spanner multi-region configurations replicate each tablet across chosen regions and let you designate a default leader region for writes. Because every write commit is coordinated by the Paxos leader and its synchronous quorum, placing that leader near your gaming company's player-facing service reduces the network distance (RTT) for leaderboard score updates. Multi-region configurations also add regional and read-only replicas, so reads can be served close to players while write latency from remote locations drops substantially.

Why this answer

Cloud Spanner's multi-region configuration is designed to reduce write latency for globally distributed users by placing write-capable replicas in multiple geographic regions. This allows writes to be committed at the nearest replica, leveraging Spanner's TrueTime and Paxos-based replication to maintain strong consistency across regions. A single-region deployment forces all writes to a single location, causing high latency for distant clients during peak hours.

Exam trap

Google Cloud often tests the misconception that scaling a database horizontally by adding nodes always reduces latency, but in a single-region Spanner setup, adding nodes only increases throughput and storage, not geographic proximity, which is the root cause of high write latency for global users.

How to eliminate wrong answers

Option B is wrong because application-level caching with Memorystore does not reduce write latency to the database; it only improves read performance for cached data, and writes still must go to the single-region Spanner instance. Option C is wrong because Cloud Bigtable is optimized for high-throughput, low-latency reads and writes for analytical workloads, but it does not support strong transactional consistency or SQL queries, making it unsuitable for a leaderboard that requires real-time, consistent updates. Option D is wrong because adding more nodes to the existing single-region Spanner instance increases throughput and storage capacity but does not reduce write latency for clients far from that region; the write path still requires consensus across replicas in the same geographic location.

525
Matchingmedium

Match each Google Cloud AI/ML term to its definition.

Drag a concept onto its matching description — or click a concept then click the description.

Concepts
Matches

Unified ML platform for building and deploying models

Train custom models with minimal code

Natural language understanding for chatbots

Image recognition and analysis

Text analysis and entity extraction

Why these pairings

Vertex AI is the unified platform; AutoML automates model training; Cloud TPUs accelerate computation. Common confusions include swapping these definitions.

Page 6

Page 7 of 12

Page 8