Cloud Digital Leader — Practice Set 1

Question 1

Which of the following best describes the difference between CapEx and OpEx in the context of cloud adoption?

Accepted Answer

CapEx (capital expenditure) involves large upfront investments in physical assets such as servers and data centers, which are then depreciated over time. OpEx (operational expenditure) refers to ongoing pay-as-you-go costs such as cloud service subscriptions. Cloud adoption shifts spending from CapEx to OpEx, providing greater financial flexibility. Options A, C, and D reverse or conflate these definitions.

Question 2

A retail company currently hosts its e-commerce platform in an on-premises data center. The company wants to move to Google Cloud to reduce infrastructure management overhead while maintaining the ability to burst to the cloud during peak shopping seasons. Which cloud migration strategy best describes this approach?

Accepted Answer

Rehosting (lift-and-shift) moves workloads to the cloud with minimal changes, enabling the company to benefit from elastic scaling during peak seasons without significant re-engineering effort. Retaining keeps workloads on-premises, retiring removes them, and re-architecting involves substantial redesign — none of which match the described scenario of reducing management overhead quickly.

Question 3

Google Cloud's global infrastructure is organized into regions and zones. What is the primary purpose of deploying resources across multiple zones within a region?

Accepted Answer

Zones are isolated locations within a region, each with independent power, cooling, and networking. Deploying resources across multiple zones within a region ensures high availability and fault tolerance — if one zone experiences an outage, resources in other zones continue to serve traffic. Multi-region deployments (not multi-zone within a region) address data residency and latency for geographically distributed users.

Question 4

A manufacturing company is evaluating whether to move its ERP system to Google Cloud or keep it on-premises. The CIO cites concerns about losing control of proprietary data and unpredictable monthly costs. Which cloud characteristic directly addresses the concern about unpredictable costs?

Accepted Answer

Cloud elasticity means resources scale up or down based on demand; while costs vary with usage, Google Cloud provides budget alerts and quotas to help organizations forecast and control spending. Multi-tenancy reduces vendor costs but doesn't directly address unpredictable customer billing. Google Cloud does not use exclusively open-source software, and colocation describes a different model (customer-owned hardware in a third-party facility) not applicable here.

Question 5

A business analyst needs to run ad hoc SQL queries on petabytes of structured sales data without managing any servers or clusters. Which Google Cloud service is best suited for this requirement?

Accepted Answer

BigQuery is Google Cloud's fully managed, serverless data warehouse designed for large-scale analytical SQL queries across petabytes of data. Cloud Spanner is a globally distributed relational database for transactional workloads. Cloud SQL is a managed relational database for smaller transactional workloads. Firestore is a NoSQL document database unsuitable for complex analytical SQL queries at petabyte scale.

Question 6

A media company wants to ingest and process millions of events per second from user clickstreams in real time, and then store the processed results for long-term analysis. Which combination of Google Cloud services should they use?

Accepted Answer

Pub/Sub is Google Cloud's managed messaging service for high-throughput event ingestion. Dataflow is a fully managed stream and batch processing service based on Apache Beam, ideal for real-time data transformation. BigQuery provides the analytical storage and querying capability for processed results. Cloud Storage is not designed for real-time event ingestion, and the other combinations do not provide the right services for this streaming pipeline pattern.

Question 7

An organization wants to store raw, unstructured data from IoT sensors, log files, and images in a central repository for future processing and analytics. What Google Cloud concept best describes this storage pattern?

Accepted Answer

A data lake stores raw, unstructured, semi-structured, and structured data in its native format at scale, enabling diverse future analytics use cases. Google Cloud Storage is commonly used to build data lakes. A data warehouse (like BigQuery) stores structured, processed data optimized for SQL queries. An operational database handles transactional workloads, and a data mart is a focused subset of a warehouse — neither matches the IoT/log/image raw storage pattern.

Question 8

A retail chain wants to predict which customers are likely to churn in the next 30 days based on purchase history and browsing behavior. They have a data science team but limited ML infrastructure. Which Google Cloud approach best balances speed-to-production with model customization?

Accepted Answer

Vertex AI is Google Cloud's unified ML platform that provides managed infrastructure for training custom models, running experiments, and deploying models to production — ideal for a data science team that needs customization without managing raw infrastructure. Cloud Vision API is for image recognition, not churn prediction. BigQuery SQL alone cannot build predictive ML models. An on-premises Hadoop cluster contradicts the goal of leveraging cloud infrastructure.

Question 9

A business intelligence team needs to create interactive dashboards and reports that connect directly to BigQuery without exporting data. Which Google Cloud service provides this capability?

Accepted Answer

Looker is Google Cloud's business intelligence and data analytics platform that integrates natively with BigQuery to enable interactive dashboards, reports, and data exploration without requiring data exports. Dataflow is a data processing service, Cloud Composer is a managed Apache Airflow workflow orchestration service, and Pub/Sub is a messaging service — none provide BI dashboard functionality.

Question 10

A startup wants to deploy a containerized web application on Google Cloud without managing the underlying virtual machines or Kubernetes clusters. Which Google Cloud compute service is the best fit?

Accepted Answer

Cloud Run is a fully managed serverless compute platform that runs stateless containers without requiring teams to manage VMs, nodes, or Kubernetes clusters. Compute Engine requires VM management, GKE requires cluster configuration and management, and Bare Metal Solution provides raw hardware — all involve more infrastructure management than Cloud Run.

Question 11

A financial services company is modernizing a legacy monolithic banking application. Their architecture team recommends breaking it into independently deployable services that communicate over APIs. What architectural pattern does this describe?

Accepted Answer

Microservices architecture involves decomposing a monolithic application into small, loosely coupled services that are independently deployable, scalable, and maintainable. Each service communicates over APIs. Lift-and-shift moves the monolith as-is to the cloud. Re-platforming involves moving to a managed service with minimal code changes but doesn't decompose the architecture. Colocation is a hosting model, not an architectural pattern.

Question 12

A company runs a stateful legacy database application on-premises. The migration team has assessed that rewriting the application is too costly. The team wants to move to Google Cloud with the least code changes while still benefiting from managed infrastructure. Which strategy is most appropriate?

Accepted Answer

Re-platforming (also called 'lift and optimize') involves moving an application to a managed cloud service with minimal code changes, allowing the company to benefit from managed infrastructure (patching, backups, high availability) without a full rewrite. Re-architecting requires significant effort to rebuild. Retaining keeps costs and operational burden on-premises. Retiring assumes a suitable SaaS replacement exists, which the question does not indicate.

Question 13

An organization needs to run containers at scale with support for automatic load balancing, self-healing, and rolling updates. They want to maintain control over the orchestration configuration. Which Google Cloud service should they use?

Accepted Answer

GKE is Google Cloud's managed Kubernetes service that provides full control over container orchestration, including load balancing, self-healing (automatic pod restarts), rolling updates, and scaling. App Engine abstracts away container orchestration, Cloud Functions is for event-driven serverless code, and Cloud Storage stores objects — none provide the orchestration control that Kubernetes/GKE offers.

Question 14

A company wants to modernize its infrastructure and application stack. They plan to move first to cloud VMs, then to containers, and eventually to serverless. What does this progression represent?

Accepted Answer

Moving from VMs (IaaS) to containers (CaaS) to serverless (PaaS/FaaS) represents an incremental modernization journey where each step abstracts more infrastructure management from the team. This is a common cloud modernization pattern Google Cloud supports through Compute Engine, GKE, and Cloud Run/Cloud Functions respectively. It is not a hybrid, multi-cloud, or data migration strategy.

Question 15

Under the Google Cloud shared responsibility model, which of the following is always the customer's responsibility regardless of the service type (IaaS, PaaS, or SaaS)?

Accepted Answer

Under the shared responsibility model, customers are always responsible for managing who has access to their resources — including IAM policies, user account management, and authentication settings. Google manages physical security, hypervisor patching, and network hardware across all service models. The customer's security responsibilities shift depending on the service type, but access management remains constant.

Question 16

A healthcare company needs to ensure that sensitive patient data stored in Google Cloud Storage is automatically identified and redacted before sharing with external partners. Which Google Cloud service should they use?

Accepted Answer

Cloud Data Loss Prevention (DLP) is a fully managed service that discovers, classifies, and can redact or mask sensitive data such as personally identifiable information (PII) and protected health information (PHI). Cloud Armor focuses on network-level threat protection, VPC Service Controls restricts data movement between services, and IAP controls application access — none perform data content scanning and redaction.

Question 17

A company's security team needs a comprehensive record of who did what, when, and where across all Google Cloud services in their organization — including administrative actions and data access events. Which Google Cloud capability provides this?

Accepted Answer

Cloud Audit Logs provide an immutable record of activity across Google Cloud services, including Admin Activity logs (who changed what configuration), Data Access logs (who read or modified data), and System Event logs. This is essential for compliance, forensics, and security investigation. Cloud Monitoring tracks performance metrics, Cloud Trace analyzes request latency, and Error Reporting surfaces application errors — none provide the administrative audit trail.

Question 18

An organization wants to protect its public-facing web applications from SQL injection, cross-site scripting (XSS), and distributed denial-of-service (DDoS) attacks. Which Google Cloud service should they deploy?

Accepted Answer

Cloud Armor is Google Cloud's web application firewall (WAF) and DDoS protection service. It provides pre-configured and custom rules to block common web attacks including SQL injection and XSS, as well as volumetric DDoS protection. Cloud VPN encrypts network connections, Cloud CDN accelerates content delivery, and Cloud NAT manages outbound internet access — none provide WAF or DDoS protection for web applications.

Question 19

A company's application experiences high traffic on weekdays and very low traffic on weekends. They want to minimize costs while ensuring consistent performance. Which Google Cloud capability allows the application to automatically adjust the number of running instances based on demand?

Accepted Answer

Autoscaling automatically adjusts the number of compute instances based on demand metrics such as CPU utilization or custom metrics. This ensures performance during peak traffic (weekdays) while reducing instance count during low-traffic periods (weekends) to minimize costs. Committed use discounts reduce prices but require fixed usage commitments. Preemptible VMs are low-cost but can be interrupted. Sustained use discounts apply based on monthly usage, not real-time demand.

Question 20

A team is setting up operational targets for a customer-facing API. They want to define the target reliability level for the API and create dashboards to track whether they are meeting it. What Google Cloud operations concept should they implement?

Accepted Answer

A Service Level Objective (SLO) is an internal reliability target (e.g., 99.9% availability) defined for a service and tracked using metrics dashboards. Cloud Monitoring in Google Cloud supports SLO creation and tracking. Error budgets are derived from SLOs and measure remaining unreliability budget, but are not the primary concept for setting the target. An SLA is an external contractual commitment. Cloud Logging collects log data but does not define reliability targets.

Question 21

A large enterprise wants to organize its Google Cloud resources so that different departments have isolated billing, separate IAM policies, and centralized governance by the IT team. Which Google Cloud resource hierarchy structure supports this requirement?

Accepted Answer

Google Cloud's resource hierarchy — Organization > Folders > Projects > Resources — enables centralized governance via the organization node while allowing department-level isolation through folders and project-level billing separation. IAM policies are inherited down the hierarchy, enabling centralized policy enforcement with delegated administration. A single project cannot isolate billing or IAM by department. Separate Google Cloud accounts would prevent centralized governance. Firewall rules control network traffic, not resource organization or billing.

Question 22

A company is building a loan approval model using machine learning. During evaluation, they discover the model has significantly lower accuracy for applicants from certain demographic groups. Which Responsible AI principle is most relevant to this concern?

Accepted Answer

Fairness in AI means that systems should not produce discriminatory outcomes or systematically disadvantage certain groups. A loan approval model that performs worse for specific demographic groups violates the fairness principle by potentially amplifying historical biases in training data. Privacy protects data, reliability ensures consistent performance, and explainability helps humans understand model decisions — none directly address differential performance across demographic groups.

Question 23

A customer service team wants to automatically analyze thousands of customer support emails and categorize them by sentiment (positive, negative, neutral). Which Google Cloud pre-trained API should they use?

Accepted Answer

The Cloud Natural Language API provides pre-trained models for text analysis including sentiment analysis (positive/negative/neutral), entity recognition, and syntactic analysis. This is purpose-built for analyzing written text such as emails. Cloud Vision API works with images, Cloud Translation API converts text between languages, and Cloud Speech-to-Text converts audio to text — none perform sentiment analysis on written text.

Question 24

An AI team at a healthcare company is deploying a diagnostic support model. Leadership wants assurance that the model's decisions can be audited and that physicians understand why the model made a specific recommendation. Which Responsible AI capability in Vertex AI addresses this requirement?

Accepted Answer

Vertex Explainable AI provides feature attribution methods (such as integrated gradients and SHAP values) that show which input features most influenced a model prediction. This enables physicians to understand why the model made a specific recommendation and supports auditability. Feature Store manages reusable ML features, Pipelines orchestrates ML workflows, and Model Monitoring detects drift — none provide per-prediction explanations for model decisions.

Question 25

A company is evaluating generative AI tools for internal use. An executive asks what 'hallucination' means in the context of large language models. Which definition is correct?

Accepted Answer

Hallucination in the context of LLMs refers to the model generating plausible-sounding but factually incorrect or fabricated information — for example, inventing citations, facts, or events that don't exist. This is a key risk of generative AI that organizations must manage through grounding techniques, retrieval-augmented generation (RAG), and human review. The other options describe different phenomena: training data leakage, intentional creative writing, and repetition loops.

— Cloud Digital Leader