Google Cloud Digital Leader Practice Exam

• A. Using technology to change how it works and serves customers Correct answer
• B. Replacing on-premises servers with newer, faster hardware
• C. Scanning paper records into searchable digital archives
• D. Outsourcing its IT help desk to an external provider

Option A is correct because digital transformation means using technology to fundamentally change how an organization operates and serves customers. Hardware refreshes, digitizing paper, and outsourcing the help desk are technology changes but do not rethink operations or customer value.

• A. On-demand provisioning of resources Correct answer
• B. Replication of data across regions
• C. Automatic scaling during peak load
• D. Pay-only-for-usage billing

Option A is correct because on-demand provisioning lets the retailer get compute resources in minutes rather than waiting months for hardware purchase and installation. Replication, autoscaling and usage-based billing are cloud benefits but do not address procurement speed.

• A. Platform as a Service (PaaS)
• B. Software as a Service (SaaS) Correct answer
• C. Function as a Service (FaaS)
• D. Infrastructure as a Service (IaaS)

Option B is correct because SaaS delivers a complete application over the internet while the provider manages everything underneath it. PaaS and IaaS leave the application or more of the stack to the customer, and FaaS runs individual functions rather than finished applications.

• A. Encryption of data at rest Correct answer
• B. Encryption of data in transit Correct answer
• C. Customer-managed encryption keys for all data
• D. Replication of data to every region
• E. Dedicated physical servers per customer

Options A and B are correct because Google Cloud encrypts customer data at rest and in transit by default, with no extra configuration. Customer-managed keys are optional, data is not replicated to every region, and customers share infrastructure rather than receiving dedicated servers.

• A. Relational data
• B. Structured data
• C. Semi-structured data
• D. Unstructured data Correct answer

Option D is correct because unstructured data such as photos, audio and video has no predefined data model or schema. Relational and structured data follow a fixed schema, and semi-structured data (like JSON) carries some organising tags.

• A. Scaling web servers automatically during seasonal traffic peaks
• B. Serving store pages from locations closer to shoppers
• C. Forecasting demand from past sales and seasonal trends Correct answer
• D. Shifting infrastructure spending from capital to operating expense

Option C is correct because analysing its own historical sales and seasonal trends lets the retailer forecast demand and order stock on evidence rather than instinct. The other options are cloud benefits about scaling, latency and cost model, not about data-driven decisions.

• A. Establishing a data governance program Correct answer
• B. Replicating databases across several regions
• C. Consolidating its sources into a data warehouse
• D. Moving older records to a lower-cost storage class

Option A is correct because data governance defines agreed definitions, ownership, quality and access rules so data can be trusted. Replication, a warehouse and storage tiering are technical choices that do not by themselves establish ownership or policy.

• A. A patient's home address Correct answer
• B. A customer's passport number Correct answer
• C. Average basket size by region
• D. Peak CPU load on a web server
• E. A product's catalog number

Options A and B are correct because a home address and a passport number each identify a specific individual, which is what makes data personally identifiable. Average basket size by region, server CPU load and a product catalog number describe aggregates, systems or products, not a person.

• A. Inference Correct answer
• B. Feature engineering
• C. Training
• D. Labeling

Option A is correct because inference is applying an already-trained model to new data to produce a prediction. Training builds the model, feature engineering prepares inputs, and labeling tags data, so none of those is the prediction step.

• A. Reinforcement learning
• B. Supervised learning Correct answer
• C. Dimensionality reduction
• D. Unsupervised learning

Option B is correct because the records carry known outcomes (cancelled or not), which are labels, so supervised learning can learn to predict them. Unsupervised learning and dimensionality reduction work without labels, and reinforcement learning learns from rewards in an environment.

• A. A model corrects errors in its training data automatically
• B. Data quality affects storage cost but not accuracy
• C. Poor data affects supervised models, not unsupervised ones
• D. A model's accuracy depends on the data it learned from Correct answer

Option D is correct because a model can only learn patterns present in its training data, so poor or biased data leads to poor predictions. Models do not fix bad data automatically, data quality affects accuracy rather than just storage cost, and it matters for unsupervised models as well.

• A. Replacing it with a SaaS product
• B. Refactoring into microservices
• C. Rehosting (lift and shift) Correct answer
• D. Rebuilding on serverless services

Option C is correct because rehosting (lift and shift) moves an application as-is onto cloud infrastructure with little or no change. Replacing with SaaS, refactoring into microservices and rebuilding on serverless all change the application itself.

• A. Compute Engine
• B. Migrate to Virtual Machines
• C. Google Kubernetes Engine
• D. Google Cloud VMware Engine Correct answer

Option D is correct because Google Cloud VMware Engine runs VMware workloads natively on the familiar VMware tooling. Compute Engine and GKE need the workloads converted, and Migrate to Virtual Machines converts VMs to Compute Engine rather than keeping VMware tools.

• A. GKE Enterprise Correct answer
• B. Cloud Run
• C. Compute Engine managed instance groups
• D. App Engine standard environment

Option A is correct because GKE Enterprise manages clusters consistently across Google Cloud and on-premises environments from one place. Cloud Run, managed instance groups and App Engine standard do not give a unified multi-environment cluster management plane.

• A. Maintaining the hardware that runs the servers
• B. Controlling which users can access its data Correct answer
• C. Securing the physical data center facilities
• D. Operating the network between data centers

Option B is correct because under the shared responsibility model Google secures the hardware, facilities and network, while the customer remains responsible for deciding who can access its data. Options A, C and D are infrastructure duties that Google handles.

• A. Shared fate
• B. Separation of duties
• C. Least privilege Correct answer
• D. Defense in depth

Option C is correct because least privilege grants each person only the permissions needed for their job. Separation of duties splits critical tasks among people, defense in depth layers controls, and shared fate is Google's partnership model with customers.

• A. Encryption of data at rest
• B. A longer minimum password length
• C. A web application firewall
• D. Multi-factor authentication Correct answer

Option D is correct because multi-factor authentication requires a second factor beyond the password, so a phished password alone is not enough to sign in. Encryption at rest, a longer password and a web application firewall do not stop sign-in with a stolen valid password.

• A. Cloud Logging log sinks
• B. Resource Manager folders
• C. Cloud Monitoring uptime checks
• D. Cloud Billing budget alerts Correct answer

Option D is correct because Cloud Billing budget alerts notify people by email when spending reaches thresholds set as a percentage of a budget amount. Log sinks route logs, folders organize resources, and uptime checks test availability, so none of them track cost.

• A. It is billed per employee who holds an account
• B. It commits to fixed monthly capacity whether used or not
• C. It buys servers up front and writes them off over years
• D. It is billed for the resources it consumes Correct answer

Option D is correct because pay-as-you-go means the customer is billed only for the resources it actually consumes. Fixed capacity commitments, up-front server purchases, and per-employee billing describe other models, not cloud consumption pricing.

• A. Free trial credits
• B. Committed use discounts Correct answer
• C. Spot VM pricing
• D. Sustained use discounts

Option B is correct because committed use discounts give a lower rate in exchange for committing to one or three years of usage. Sustained use discounts apply automatically with no commitment, Spot VMs are cheap but preemptible, and free credits are not a term-based discount.