SaveMyCert
Log in
5 of 5 free questions left today·for 10 a day
ACE · Domain 2

Planning and implementing a cloud solution practice questions

Planning and implementing a cloud solution is worth 30% of the ACE exam — the heaviest of the 4 domains. Planning and deploying compute, storage/data, networking, and resources via tooling. Official (approximate) weighting ~30%. 6 fully worked examples are further down this page, answers included.

Exam weight
30%
the heaviest of the 4 domains
Questions
144
across 4 topics
Free, no account
5/day
a free account allows 10 a day
Explanations
Every option
right and wrong

Build a practice session

5 free questions left today.

Domains

How many?

5 free questions left today.

Mode

Ready when you are

10 fresh questions drawn across 1 of 4 domains, in Learn mode.

Review what you missed

Every question you answer incorrectly, and every question you flag while practising, is saved here automatically. Finish a session and you can come back to re-drill just those.

6 sample Planning and implementing a cloud solution questions, fully explained

Questions from the ACE bank mapped to domain 2, with the answer key and the reasoning behind every option. None of them repeat the examples on the main ACE practice page.

Question 1Planning and implementing a cloud solution

A media company runs a rendering job on Compute Engine for about three hours each night. The job checkpoints its progress to Cloud Storage every few minutes and can resume from the last checkpoint if a worker VM stops. The company wants to lower the cost of the job without changing its code. What should you do?

Choose one.

  • a
    Run the rendering workers as Spot VMs Correct

    Spot VMs carry the deepest Compute Engine discount in exchange for possible preemption, which this checkpointed, resumable job tolerates by design.

  • b
    Run the rendering workers to earn sustained use discounts

    Sustained use discounts apply automatically only once a VM runs for more than about 25% of the month; three hours a night is roughly 12% of the month, so the job earns little or nothing.

  • c
    Run the rendering workers under a 1-year committed use discount

    Committed use discounts bill the committed capacity around the clock, which suits steady 24/7 usage, not a job that runs three hours a night.

  • d
    Run the rendering workers on sole-tenant nodes

    Sole-tenant nodes dedicate a physical host for isolation or licensing reasons and cost more than shared capacity.

The concept

Spot VMs are spare Compute Engine capacity sold at a steep discount (typically 60-91% off on-demand) that Compute Engine can preempt at any time; they are the standard cost lever for fault-tolerant, interruptible work.

Why that’s the answer

Because the job checkpoints every few minutes, a preemption costs only a little rework, so Spot pricing fits it exactly. A committed use discount is tempting, but you pay for the commitment 24 hours a day while the job runs three. Sustained use discounts only start once a VM runs for more than a quarter of the month, which a three-hour nightly job never reaches, and sole-tenant nodes add a premium.

How to reason it out
  1. Confirm the workload tolerates interruption: it checkpoints and resumes.
  2. Note the usage pattern: about three hours a night, roughly 12% of the month.
  3. Rule out discounts tied to steady usage: committed use (billed 24/7) and sustained use (needs more than 25% of the month).
  4. Match interruptible, short-duration work to Spot VMs.

Exam tip: Interruptible, checkpointed work points to Spot VMs; committed use and sustained use discounts only pay off for steady usage.

Choosing and Deploying Compute: Compute Engine, GKE, and Cloud Run — the lesson that teaches this.

Question 2Planning and implementing a cloud solution

An in-memory analytics service will run on one N2 VM. Profiling shows it needs 6 vCPUs and 52 GB of memory in a single address space, and it never uses more than 6 vCPUs. You want to run it at the lowest cost without changing the application. Which machine configuration should you choose?

Choose one.

  • a
    Create an N2 custom machine type with 6 vCPUs and 52 GB, using extended memory Correct

    A custom machine type sets the exact shape; 52 GB over 6 vCPUs is above the N2 per-vCPU memory ratio, so extended memory supplies the difference and you pay only for what the workload needs.

  • b
    Use the n2-highmem-8 predefined machine type, which provides 8 vCPUs and 64 GB

    n2-highmem-8 meets the memory need but buys 2 extra vCPUs and 12 GB of extra memory that the service never uses.

  • c
    Use two n2-highmem-4 VMs, which have 4 vCPUs and 32 GB each

    The data must sit in one address space, so splitting it over two VMs would require changing the application.

  • d
    Use an n2-standard-8 VM and add a swap file on an SSD Persistent Disk

    n2-standard-8 has 32 GB; swapping the rest of an in-memory dataset to disk would cripple performance, and it still pays for 8 vCPUs.

The concept

Custom machine types decouple vCPU count from memory size. Each family has a default memory-per-vCPU range; extended memory lets a custom shape exceed it, billed per extra GB.

Why that’s the answer

The 6 vCPU / 52 GB shape falls between predefined types, which is the gap custom machine types fill, and because 52 GB is more than the N2 per-vCPU ratio allows for 6 vCPUs, extended memory covers the rest. n2-highmem-8 is the tempting choice because it works, but it overprovisions both vCPUs and memory. Two VMs break the single-address-space constraint, and swap to disk destroys in-memory performance.

How to reason it out
  1. Write down the exact shape needed: 6 vCPUs, 52 GB, one VM.
  2. Check predefined types: each either lacks memory or adds unused vCPUs.
  3. Use a custom machine type, adding extended memory when the memory-per-vCPU ratio exceeds the family default.

Exam tip: When predefined shapes force you to overprovision, use a custom machine type, with extended memory if the memory-to-vCPU ratio is high.

Choosing and Deploying Compute: Compute Engine, GKE, and Cloud Run — the lesson that teaches this.

Question 3Planning and implementing a cloud solution

Your web tier runs in a regional managed instance group that autoscales on CPU utilization. Load tests show the e2-standard-2 VMs are undersized, and you want the group to run e2-standard-4 VMs from now on, including VMs the autoscaler adds later, with no outage and the least manual effort. What should you do?

Choose one.

  • a
    Edit the machine type in the group's current instance template

    Instance templates are immutable; you cannot edit an existing template's machine type, you can only create a new template.

  • b
    Stop each VM and run set-machine-type on it individually

    Changing VMs by hand is manual, and the group would keep creating e2-standard-2 VMs from its unchanged template whenever it scales out or recreates an instance.

  • c
    Create a new instance template and roll it out to the group Correct

    Creating a new template with e2-standard-4 and starting a rolling update replaces the VMs gradually without an outage, and every VM added later comes from the new template.

  • d
    Delete the group and recreate it from a new instance template

    Recreating the group would also switch the machine type, but deleting it takes the web tier offline, which breaks the no-outage requirement.

The concept

A managed instance group creates every VM from its instance template. Templates are immutable, so changing the fleet means creating a new template and updating the group to it, usually with a rolling update.

Why that’s the answer

The change must persist for VMs the autoscaler creates later, so it has to live in the template the group uses. Because templates cannot be edited, you create a new one and roll it out; a rolling update replaces VMs a few at a time with no outage. Editing VMs by hand is undone the next time the group creates a VM, and recreating the group works but causes an outage.

How to reason it out
  1. Recognise that the group stamps every VM, including future ones, from its instance template.
  2. Recall that instance templates are immutable.
  3. Create a new template with e2-standard-4.
  4. Start a rolling update of the group to the new template (gcloud compute instance-groups managed rolling-action start-update).

Exam tip: To change a managed instance group's VMs, create a new instance template and roll it out; templates are never edited in place.

Choosing and Deploying Compute: Compute Engine, GKE, and Cloud Run — the lesson that teaches this.

Question 4Planning and implementing a cloud solution

A platform team is moving 30 microservices to Google Cloud. They deploy with Kubernetes manifests, Helm charts and a custom operator, and they have no staff for node upgrades, OS patching or capacity planning. Finance wants the bill to follow what the workloads ask for rather than idle VM capacity. Which deployment option should they choose?

Choose one.

  • a
    Deploy the microservices to a GKE Standard cluster with node auto-provisioning

    Node auto-provisioning creates node pools automatically, but in Standard mode the team still owns node configuration and maintenance decisions, and billing remains per node VM, so idle capacity is still paid for.

  • b
    Deploy the microservices as Cloud Run services from the container images

    Cloud Run removes node management, but it is not a Kubernetes cluster: Helm charts and a custom operator (custom resources and controllers) have nothing to run against.

  • c
    Deploy the microservices to managed instance groups of container-optimized VMs

    MIGs run containers on VMs with no Kubernetes control plane, so the manifests, charts and operator stop working, and the team gains VM operations work and per-VM billing.

  • d
    Deploy the microservices to a GKE Autopilot cluster with default settings Correct

    Autopilot keeps the Kubernetes API (so manifests, Helm and operators work), has Google provision, upgrade and secure the nodes, and bills for the CPU, memory and storage the Pods request.

The concept

GKE offers two modes: Standard, where you manage node pools and pay per node, and Autopilot, where Google manages the nodes and you pay for the CPU, memory, and storage your Pods request.

Why that’s the answer

Helm charts and a custom operator need a real Kubernetes API; no node staff and request-based billing describe Autopilot. Standard with node auto-provisioning automates node pool creation but leaves node lifecycle and per-node billing with the team. Cloud Run and MIGs drop the Kubernetes API that the manifests, charts and operator depend on.

How to reason it out
  1. Confirm the team needs real Kubernetes: manifests, Helm and a custom operator.
  2. Note the operating constraint: nobody to manage nodes.
  3. Note the billing constraint: pay for what workloads request, not for VMs.
  4. Select GKE Autopilot, which combines the Kubernetes API with Google-managed nodes and Pod-request billing.

Exam tip: Kubernetes API without node operations and with per-Pod billing means GKE Autopilot, not Standard mode or Cloud Run.

Choosing and Deploying Compute: Compute Engine, GKE, and Cloud Run — the lesson that teaches this.

Question 5Planning and implementing a cloud solution

Your company is deploying a production GKE Standard cluster. A key requirement is that the Kubernetes control plane must remain available even if a single zone experiences an outage. Which cluster configuration should you choose?

Choose one.

  • a
    A multi-zonal cluster

    A multi-zonal cluster spreads nodes across zones but runs only a single control plane replica in one zone - if that zone fails, the Kubernetes API becomes unavailable even though workloads keep running.

  • b
    A zonal cluster with node auto-repair enabled

    Node auto-repair replaces unhealthy nodes; it does nothing for the control plane, which in a zonal cluster lives in exactly one zone.

  • c
    A regional cluster Correct

    A regional cluster replicates the control plane across multiple zones in the region, so the Kubernetes API stays available during a zonal outage.

  • d
    A private cluster

    Private clusters remove public IPs from nodes for network security; the private setting is independent of control-plane replication and does not add zonal resilience.

The concept

GKE cluster availability is determined by where the control plane runs: zonal and multi-zonal clusters have a single-zone control plane, while regional clusters replicate the control plane across zones.

Why that’s the answer

Only a regional cluster satisfies the requirement, because it runs multiple control-plane replicas in different zones of the region. The classic trap is the multi-zonal cluster: its nodes span zones, but its control plane does not, so a zonal outage can take down the API server. Auto-repair and private networking address different concerns entirely.

How to reason it out
  1. Identify that the requirement is about control-plane availability, not just node availability.
  2. Recall that zonal and multi-zonal clusters both run the control plane in a single zone.
  3. Choose a regional cluster, which replicates both the control plane and (by default) nodes across multiple zones.

Exam tip: Multi-zonal spreads only nodes; regional is the only GKE configuration that makes the control plane survive a zone outage.

Choosing and Deploying Compute: Compute Engine, GKE, and Cloud Run — the lesson that teaches this.

Question 6Planning and implementing a cloud solution

You run a GKE cluster with private nodes, so the nodes have only internal IP addresses. During startup, Pods must download packages from public internet repositories such as PyPI, but nothing on the internet may open connections to the nodes. What should you configure?

Choose one.

  • a
    Add an authorized network for the cluster's control-plane endpoint

    Authorized networks restrict who can reach the control-plane endpoint; they give the nodes no path out to the internet.

  • b
    Configure Cloud NAT for the cluster's subnet and region Correct

    Cloud NAT gives resources without external IPs outbound-only internet access, so the nodes and Pods can fetch packages while no inbound connections are possible.

  • c
    Expose the Pods through a Service of type LoadBalancer

    A LoadBalancer Service handles inbound traffic to the Pods; it does not give the nodes an egress path, and it adds inbound exposure.

  • d
    Enable Private Google Access on the cluster's subnet

    Private Google Access lets internal-only VMs reach Google APIs and services, not third-party internet repositories such as PyPI.

The concept

Private nodes have internal IPs only. Outbound internet access comes from Cloud NAT (configured on a Cloud Router in the region), which performs source NAT for egress without allowing inbound connections.

Why that’s the answer

Cloud NAT is the egress path for internal-only nodes, and it accepts no unsolicited inbound connections. Private Google Access is the tempting near-twin, but it only reaches Google APIs and services. Authorized networks govern access to the control plane, and a LoadBalancer Service is inbound.

How to reason it out
  1. Identify the need: outbound internet access from nodes with no external IPs, no inbound exposure.
  2. Separate Google-API access (Private Google Access) from general internet access (Cloud NAT).
  3. Create a Cloud Router and Cloud NAT gateway covering the cluster's subnet, including Pod ranges.

Exam tip: Private nodes reaching the public internet need Cloud NAT; Private Google Access only reaches Google APIs.

Choosing and Deploying Compute: Compute Engine, GKE, and Cloud Run — the lesson that teaches this.

What ACE domain 2 tests, topic by topic

The official exam guide breaks Planning and implementing a cloud solution into 4 topics. The question bank follows the same split, so a weak topic shows up as a cluster of misses you can go back and read.

Published ACE practice questions per topic in Planning and implementing a cloud solution
TopicWhat it coversQuestions
Planning and implementing compute resourcesOfficial ACE exam-guide sub-section. Selecting compute for a workload (Compute Engine, GKE, Cloud Run, Cloud Run functions, Agent Runtime on Gemini Enterprise Agent Platform); launching instances (availability policy, SSH keys); choosing Compute Engine storage (zonal/regional Persistent Disk, Hyperdisk); autoscaled managed instance groups with instance templates; OS Login; VM Manager; Spot VMs and custom machine types; kubectl; deploying GKE clusters (Autopilot, regional, private); deploying containerized apps to GKE; serverless deployments and event processing (Pub/Sub, Cloud Storage notifications, Eventarc); choosing GPUs vs TPUs.36
Planning and implementing storage and data solutionsOfficial ACE exam-guide sub-section. Choosing and deploying data products (Cloud SQL, BigQuery, Firestore, Spanner, Bigtable, AlloyDB, Dataflow, Pub/Sub, Managed Service for Apache Kafka, Memorystore); choosing and deploying storage products (Cloud Storage, Filestore, NetApp Volumes, Managed Lustre) and Cloud Storage classes (Standard, Nearline, Coldline, Archive); loading data (CLI upload, from Cloud Storage, Storage Transfer Service); maintaining multi-region redundancy.36
Planning and implementing networking resourcesOfficial ACE exam-guide sub-section. Creating a VPC with subnets (custom mode, Shared VPC, VPC Network Peering); creating and applying VPC firewall rules and Cloud NGFW policies with ingress/egress attributes; using Tags (secure Tags) and service accounts in NGFW policy rules; establishing connectivity (Cloud VPN, VPC Network Peering, Cloud Interconnect); choosing and deploying load balancers; differentiating Network Service Tiers.36
Planning and implementing resources using toolingOfficial ACE exam-guide sub-section. Infrastructure as Code tooling (Fabric FAST, Config Connector, Terraform, Helm); AI-assisted planning and implementation (Gemini CLI, Google Antigravity, Gemini Cloud Assist, Application Design Center).36
Total144

Revise Planning and implementing a cloud solution before you drill it

Other ACE domains

Planning and implementing a cloud solution: your questions

Planning and implementing a cloud solution is domain 2 of the ACE exam guide and carries 30% of the scored content — the heaviest of the 4 domains. On a 55-question paper that works out to roughly 17 questions, though Google Cloud does not publish an exact per-domain count and individual exam forms vary.

Source

The domain weight and topic list on this page come from the official ACE exam guide.