10 free sample questions from a bank of 287, with the correct answers and explanations. No signup required — start practising right now.
1You are migrating workloads to the cloud. The goal of the migration is to serve customers worldwide as quickly as possible According to local regulations, certain
data is required to be stored in a specific geographic area, and it can be served worldwide. You need to design the architecture and deployment for your
workloads.
What should you do?
Select a public cloud provider that is only active in the required geographic area
Select a private cloud provider that globally replicates data storage for fast data access
Select a public cloud provider that guarantees data location in the required geographic area
Select a private cloud provider that is only active in the required geographic area
Answer: C
The short version
C — Pick a public cloud that guarantees data location. That satisfies worldwide serving plus legal residency by keeping data pinned while serving it globally.
Key concepts in this question
Data residency: regulation demanding storage in a named geography even when access is worldwide
Public cloud regions: providers with declared locations and controls to pin data at rest
Private single-area clouds: fail global speed goals and scale compared with global public footprints
Global replication: useful for speed but must respect residency guarantees rather than copy restricted data everywhere
Why C is correct
Serving customers fast worldwide while obeying location rules requires a provider that contractually and technically enforces storage location. A public provider with region pinning plus global edge serving meets both speed and compliance. Limiting to only that area or using private clouds sacrifices either reach or residency assurance.
Why the others are wrong
A. limits presence to one area, harming worldwide latency and availability
B. replicates storage globally without a residency guarantee, risking violation
D. confines deployment locally, defeating fast worldwide service
CDL exam tip
Residency plus worldwide speed equals public region pinning: guarantee where data rests, serve from edge, never copy restricted data blindly.
2Your organization needs a large amount of extra computing power within the next two weeks.
After those two weeks, the need for the additional resources will end.
Which is the most cost-effective approach?
Use a committed use discount to reserve a very powerful virtual machine
Purchase one very powerful physical computer
Start a very powerful virtual machine without using a committed use discount
Purchase multiple physical computers and scale workload across them
Answer: C
The short version
C — Use on-demand VMs for a two-week burst. Short-lived extra power is cheapest without commitments or hardware purchases.
Key concepts in this question
Ephemeral demand: two weeks of extra compute that then ends, favoring pay-per-use
Committed discounts: reward steady long-term use, wasteful for a brief spike
Physical purchases: incur capital cost, lead time, and idle hardware after the peak
On-demand VMs: spin up large shapes quickly and stop billing when deleted
Why C is correct
A very powerful VM without a commitment matches cost exactly to the two-week window and vanishes afterward. Committing or buying hardware spreads cost beyond need and adds procurement delay. FinOps for cloud bursts therefore prefers elastic on-demand capacity. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
A. locks in discounts suited to stable workloads, overpaying for temporary use
B. buys depreciating hardware for a transient need with setup delay
D. multiplies hardware cost and orchestration for a short-lived requirement
CDL exam tip
Two-week spike then done equals on-demand: never buy boxes or commit for transient bursts. Always map the constraint to the managed control that inherits everywhere.
3Your organization needs to plan its cloud infrastructure expenditures.
Which should your organization do?
Review cloud resource costs frequently, because costs change often based on use
Review cloud resource costs annually as part of planning your organizationג€™s overall budget
If your organization uses only cloud resources, infrastructure costs are no longer part of your overall budget
Involve fewer people in cloud resource planning than your organization did for on-premises resource planning
Answer: A
The short version
A — Review cloud costs frequently based on use. Consumption pricing shifts with usage and rates, so frequent review keeps budgets accurate.
Key concepts in this question
Variable cloud spend: metered by seconds, storage, and egress, changing as workloads scale
Frequent review: catches drift, idle resources, and price changes early for action
Annual-only planning: misses monthly variance and new service pricing in cloud
Planning ownership: cloud needs broader accountability, not fewer eyes on spend
Why A is correct
Unlike fixed on-premises depreciation, cloud bills move with every deployment choice and provider update. Regular cost reviews tie spending to actual consumption and expose optimization chances. That continuous FinOps habit is the core planning discipline the question tests. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
B. waits a year while usage-driven charges compound unchecked
C. wrongly removes infra costs from budgets when cloud simply makes them variable
D. reduces oversight just as shared responsibility and sprawl demand more coordination
CDL exam tip
Cloud budgeting equals frequent use-based review: variable meters need continuous eyes, not annual snapshots.
4The operating systems of some of your organization's virtual machines may have a security vulnerability.
How can your organization most effectively identify all virtual machines that do not have the latest security update?
View the Security Command Center to identify virtual machines running vulnerable disk images
View the Compliance Reports Manager to identify and download a recent PCI audit
View the Security Command Center to identify virtual machines started more than 2 weeks ago
View the Compliance Reports Manager to identify and download a recent SOC 1 audit
Answer: A
The short version
A — Use Security Command Center for vulnerable images. It inventories VM OS and package exposure to flag hosts missing patches.
Key concepts in this question
Security Command Center: Google central view of misconfigurations and vulnerabilities across assets
Vulnerable disk images: outdated OS builds that propagate risk to every VM using them
Compliance Reports Manager: provides audit certifications, not live patch posture per VM
Age heuristic: uptime alone does not prove patch state without vulnerability data
Why A is correct
Finding all VMs lacking the latest fix needs asset-level vulnerability findings tied to image versions. Security Command Center aggregates scanner results and highlights affected instances for remediation. That direct identification beats audit PDFs or age guesses. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
B. downloads PCI attestations without listing unpatched VMs
C. filters by start time, missing newly launched yet vulnerable hosts
D. supplies SOC evidence for auditors, not actionable patch lists
CDL exam tip
Which VMs are vulnerable equals Security Command Center: think findings per asset, not compliance letters or uptime.
5You are currently managing workloads running on Windows Server for which your company owns the licenses. Your workloads are only needed during working
hours, which allows you to shut down the instances during the weekend. Your Windows Server licenses are up for renewal in a month, and you want to optimize
your license cost.
What should you do?
Renew your licenses for an additional period of 3 years. Renew your licenses for an additional period of 3 years. Negotiate a cost reduction with your current hosting provider wherein infrastructure cost is reduced when workloads are not in use
Renew your licenses for an additional period of 2 years. Negotiate a cost reduction by committing to an automatic renewal of the licenses at the end of the 2 year period
Migrate the workloads to Compute Engine with a bring-your-own-license (BYOL) model
Migrate the workloads to Compute Engine with a pay-as-you-go (PAYG) model
Answer: D
The short version
D — Move Windows to pay-as-you-go for part-time use. Shutting down on weekends plus per-second billing beats renewing owned licenses for idle time.
Key concepts in this question
Working-hours pattern: nights and weekends powered off, so license and compute should follow usage
PAYG model: bundles Windows Server cost into running VM price with no long renewal
BYOL renewal: keeps paying for owned licenses even when instances are stopped
Multi-year lock: extends shelfware cost just before a cloud optimization window
Why D is correct
With renewal looming and VMs off two days weekly, including licensing in metered Compute Engine cost aligns spend with actual run hours. PAYG removes renewal negotiation and lets shutdowns save immediately. That optimization of license cost for intermittent workloads is exactly the scenario. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
A. renews for three years despite predictable idle time and cloud alternative
B. shortens but still prepays for licenses unused on weekends
C. migrates yet retains license burden instead of converting to usage-based pricing
CDL exam tip
Part-time Windows plus renewal soon equals PAYG: let shutdowns cut license cost instead of renewing idle entitlements.
6Your organization runs a distributed application in the Compute Engine virtual machines. Your organization needs redundancy, but it also needs extremely fast
communication (less than 10 milliseconds) between the parts of the application in different virtual machines.
Where should your organization locate this virtual machines?
In a single zone within a single region
In different zones within a single region
In multiple regions, using one zone per region
In multiple regions, using multiple zones per region
Answer: B
The short version
B — Place VMs in different zones of one region. That gives redundancy with single-digit millisecond links ideal for fast distributed parts.
Key concepts in this question
Zones in a region: separate failure domains connected by low-latency private fiber
Sub-10ms need: requires nearby placement without cross-region round trips
Single zone: fastest but no resilience to zonal failure
Multi-region: adds geographic resilience at the cost of tens of milliseconds
Why B is correct
Different zones in the same region balance the two goals: isolation from single-zone outages plus Google backbone latency well under 10ms. Single zone risks correlated failure while multi-region breaks the latency budget. Hence zonal distribution in one region is the settled pattern. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
A. concentrates risk in one failure domain despite meeting latency
C. stretches traffic across regions, exceeding the millisecond target
D. multiplies regions and latency while overengineering for the stated need
CDL exam tip
Redundant plus under 10ms equals multi-zone single-region: zones give resilience, regions break latency.
7An organization decides to migrate their on-premises environment to the cloud. They need to determine which resource components still need to be assigned
ownership.
Which two functions does a public cloud provider own? (Choose two.)
Hardware maintenance
Infrastructure architecture
Infrastructure deployment automation
Hardware capacity management
Fixing application security issues
Answer: A, D
The short version
AD — Provider owns hardware maintenance and capacity. Customers keep architecture, deployment, and app security while Google runs physical operations.
Key concepts in this question
Shared responsibility: provider secures and runs the physical plant, customer secures data and apps
Hardware maintenance: repairs, replacement, and datacenter ops owned by the cloud
Capacity management: pooling and provisioning physical servers invisible to tenants
Architecture and automation: customer choices for topology, code, and pipelines remain with the mover
Why AD is correct
Migrating to public cloud offloads undifferentiated physical work while ownership of design stays local. Maintenance and capacity are textbook provider duties in Google responsibility models. App security fixes and deployment automation clearly remain customer tasks, so they are excluded. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
B. designs system topology and requirements, a customer retained duty
C. builds pipelines and IaC for releases, owned by the migrating team
E. patches application vulnerabilities, firmly in customer scope
CDL exam tip
Who owns what in public cloud: provider equals racks and capacity, customer equals architecture, automation, and app fixes.
8You are a program manager within a Software as a Service (SaaS) company that offers rendering software for animation studios. Your team needs the ability to
allow scenes to be scheduled at will and to be interrupted at any time to restart later. Any individual scene rendering takes less than 12 hours to complete, and
there is no service-level agreement (SLA) for the completion time for all scenes. Results will be stored in a global Cloud Storage bucket. The compute resources
are not bound to any single geographical location. This software needs to run on Google Cloud in a cost-optimized way.
What should you do?
Deploy the application on Compute Engine using preemptible instances
Develop the application so it can run in an unmanaged instance group
Create a reservation for the minimum number of Compute Engine instances you will use
Start more instances with fewer virtual centralized processing units (vCPUs) instead of fewer instances with more vCPUs
Answer: A
The short version
A — Run interruptible renders on preemptible VMs. Schedulable scenes under 12 hours with no SLA fit low-cost capacity that can restart later.
Key concepts in this question
Preemptible instances: deeply discounted VMs that may be reclaimed, ideal for fault-tolerant batch
Interrupt and restart: scenes can pause and resume, tolerating preemption without business harm
Global bucket: decouples storage location from compute, allowing any zone to render
Reservations and tuning: suit steady or performance needs, not cost-first flexible batch
Why A is correct
The workload explicitly allows scheduling at will and interruption with restart, with short individual runtimes and no completion SLA. Preemptible VMs trade availability for major savings precisely for such batch rendering. Results to a global bucket preserve output regardless of which VM finishes. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
B. uses unmanaged groups without the cost advantage or preemption handling
C. reserves capacity at higher cost for a workload that tolerates eviction
D. tweaks vCPU shape without addressing the dominant cost lever of spare capacity
CDL exam tip
Can interrupt and restart with no SLA equals preemptible: batch rendering loves spare capacity savings.
9Your manager wants to restrict communication of all virtual machines with internet access; with resources in another network; or with a resource outside Compute
Engine. It is expected that different teams will create new folders and projects in the near future.
How would you restrict all virtual machines from having an external IP address?
Define an organization policy at the root organization node to restrict virtual machine instances from having an external IP address
Define an organization policy on all existing folders to define a constraint to restrict virtual machine instances from having an external IP address
Define an organization policy on all existing projects to restrict virtual machine instances from having an external IP address
Communicate with the different teams and agree that each time a virtual machine is created, it must be configured without an external IP address
Answer: A
The short version
A — Set org-policy at the root to block external IPs. Hierarchy inheritance then covers every future folder and project automatically.
Key concepts in this question
Organization policy: constraints like compute.vmExternalIpAccess that deny external IPs when enforced
Root node: top of hierarchy so all children inherit the restriction without per-team steps
Folder and project policies: leave gaps when new containers are created later
Manual agreement: relies on humans and drifts, failing audit and scale
Why A is correct
The goal spans all current plus future folders and projects, which only root-level enforcement guarantees. An org policy at the organization node with the external-IP constraint blocks creation of external addresses everywhere by inheritance. That centralized guardrail is the Google-recommended way to restrict internet exposure at scale. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
B. covers only existing folders, missing new ones teams will add
C. covers only existing projects, requiring endless reapplication
D. depends on voluntary compliance with no technical enforcement
CDL exam tip
All VMs everywhere plus future projects equals root org policy: inherit the deny-external-IP constraint from the top.
10Your multinational organization has servers running mission-critical workloads on its premises around the world. You want to be able to manage these workloads
consistently and centrally, and you want to stop managing infrastructure.
What should your organization do?
Migrate the workloads to a public cloud
Migrate the workloads to a central office building
Migrate the workloads to multiple local co-location facilities
Migrate the workloads to multiple local private clouds
Answer: A
The short version
A — Move to public cloud to centralize and shed infra. Global mission workloads gain consistent management while Google operates the underlying plant.
Key concepts in this question
Central management: single control plane for policy, deployment, and observability worldwide
Stop managing infra: shift hardware, patching of hypervisors, and capacity to the provider
Central office or colo: merely relocates physical burden without removing it
Private clouds: multiply operational load across sites instead of consolidating
Why A is correct
The dual goals are consistent global operations plus exiting infrastructure ownership, which is the definition of public-cloud migration. Centralized services replace per-site servers while provider SRE handles facilities and hardware. No on-premises consolidation achieves the stop-managing-infra aim. This reflects Google Cloud FinOps and reliability guidance to match service choice to workload pattern rather than fixed infrastructure.
Why the others are wrong
B. concentrates hardware in one building yet retains all infra management
C. spreads ops across colos, increasing inconsistency and upkeep
D. fragments control across private stacks, opposite of central simplicity
CDL exam tip
Consistent plus no more racks equals public cloud: centralize control plane and let the provider run the buildings.