GCP PCA · Domain 1 of 6

GCP PCA Domain 1: Designing and planning a cloud solution architecture

Section 1 carries the most weight on the Professional Cloud Architect exam. It asks you to turn business goals and technical constraints into a Google Cloud design, choose compute, storage, network and AI services, and plan a migration.

Exam weight
~25%
Official objectives
5
Estimated questions estimate
13–15
Suggested study time estimate
20–25 hrs

About 16 min read Reviewed October 2026

Overview

Section 1 starts from requirements, not products. Objective 1.1 covers business drivers such as cost, continuity, success metrics (KPI, ROI), build-buy-modify-deprecate decisions and integration patterns. Objective 1.2 covers technical needs: the Google Cloud Well-Architected Framework, high availability and failover, scalability, latency, backup and recovery, and Gemini Cloud Assist as a design aid. Objective 1.3 maps workloads to GKE, Cloud Run, Cloud Run functions or Compute Engine, picks object, file or database storage, and selects AI options such as Gemini models, Model Garden, Agent Builder and AI Hypercomputer.

Objective 1.4 is migration planning with Google Cloud Migration Center, dependency mapping, testing and license costs, and 1.5 asks how a design should adapt as business needs change. Every exam includes two of four published case studies (Altostrat Media, Cymbal Retail, EHR Healthcare, KnightMotives Automotive), and most case questions lean on this section. The exam is active, with no retirement announced.

How it's tested

Questions in this section present a company, its current environment and a set of stated requirements, then ask which design, service or migration approach fits best. Several options are usually technically possible. The winner is the one that meets every constraint with the least operational effort and no unnecessary cost.

  • Requirement to service

    A business or technical requirement such as 99.9% availability, global users or relational consistency must be matched to one Google Cloud service.

  • Trade-off judgement

    Two reasonable designs differ on cost, latency, operational burden or lock-in, and the stem hints which factor the company values most.

  • Migration sequencing

    You choose the order, method and tooling for moving workloads and data, usually with a fixed deadline such as an expiring data center lease.

Reasoning it through

A healthcare SaaS company must leave a colocation facility before its lease ends, keep 99.9% availability and run containers consistently on premises and in Google Cloud. Rewriting every service first would miss the deadline. A better plan moves containerized apps to GKE, uses GKE Enterprise features for consistent fleet management across both sites, links sites with Cloud Interconnect and migrates databases in waves.

Objectives 5

Official objectives from the Google Cloud outline, in outline order.

1.1 Designing a cloud solution infrastructure that meets business requirementsUse cases, cost, continuity, integration, data movement, trade-offs, workload disposition, success metrics.
What it means
This objective covers translating business language into architecture. It includes business use cases and product strategy, functional and non-functional requirements, business continuity, cost optimization, integration patterns, data movement, design trade-offs, deciding whether to build, buy, modify or deprecate a workload, and defining success through KPIs and ROI. Security, compliance and observability appear here as business needs too.
What to understand
Start each design by listing what the business must achieve and how success will be measured. A KPI such as checkout latency or time to onboard a partner tells you which design trade-off to favor. Workload disposition matters: some applications should move as they are, some should be replaced by SaaS or a managed service, and some should be retired. Cost optimization means right-sizing, committed use discounts and managed services that remove administration, not simply the cheapest option. Integration patterns cover APIs, event messaging with Pub/Sub and batch file transfer.
How it shows up
Expect case-study style questions where the stated business goal, such as cutting call-center cost or reducing administration, decides between two working designs.
Common confusion
Functional requirements say what a system does; non-functional ones such as availability and latency describe how well it must do it.
1.2 Designing a cloud solution infrastructure that meets technical requirementsWell-Architected Framework, availability and failover, elasticity, scalability, latency, Gemini Cloud Assist, backup.
What it means
This objective covers the technical qualities a design must have. It names the Google Cloud Well-Architected Framework, high availability and failover design, elasticity of resources, scalability to meet growth, performance and latency, Gemini Cloud Assist, and backup and recovery.
What to understand
Know the shape of availability in Google Cloud: zonal resources fail with a zone, regional resources such as regional managed instance groups, regional GKE clusters and Cloud SQL high availability survive a zone loss, and multi-region services such as Spanner multi-region configurations survive a regional event. Match recovery time objective (RTO) and recovery point objective (RPO) to backup, replication or active-active designs. Autoscaling managed instance groups, GKE autoscaling and Cloud Run scaling handle elastic demand. Gemini Cloud Assist helps design, explain and troubleshoot cloud resources but does not replace the architect's decisions.
How it shows up
Questions give an availability target, RTO, RPO or latency budget and ask for the least complex design that meets it.
Common confusion
A multi-zone deployment protects against zone failure but not against a full regional outage.
1.3 Designing network, storage, and compute resourcesHybrid and multicloud, AI/ML options, cloud-native networking, storage types, compute platform mapping.
What it means
This objective covers choosing the building blocks. It includes hybrid and multicloud integration, Google Cloud AI and ML solutions (Gemini models, Agent Builder, Model Garden, AI Hypercomputer), cloud-native networking, data processing technologies, storage types (object, file, database), mapping workloads to GKE, Cloud Run or Cloud Run functions, and compute choices such as Spot VMs, custom machine types and specialized workloads.
What to understand
Pick storage by access pattern: Cloud Storage for objects and media, Filestore for shared POSIX files, Cloud SQL or AlloyDB for PostgreSQL for relational workloads, Spanner for global relational scale, Bigtable for high-throughput wide-column data, Firestore for document data and BigQuery for analytics. Pick compute by control and ops effort: Compute Engine for full VM control, GKE for containers needing orchestration, Cloud Run for stateless containers that scale to zero, Cloud Run functions for event handlers. For AI, decide between calling a Gemini model, picking an open or partner model from Model Garden, building an agent, or training on AI Hypercomputer GPUs or TPUs.
How it shows up
Scenario questions describe a workload's data shape, traffic and team skills and ask which compute, storage or AI service fits.
Common confusion
Bigtable is not a relational database and does not support SQL joins or multi-row transactions.
1.4 Creating a migration plan (i.e., documents and architectural diagrams)Integration with existing systems, Migration Center assessment, methodology, testing, dependencies, licensing.
What it means
This objective covers planning a move to Google Cloud. It includes integrating with systems that stay behind, assessing and migrating systems and data with Google Cloud Migration Center, choosing a migration methodology, testing workloads, planning network and dependency needs, and estimating software license and financial impact.
What to understand
Migration Center discovers on-premises assets, groups dependencies and produces total cost of ownership estimates, which feed a wave plan. Common approaches are rehost (lift and shift), replatform (lift and optimize) and refactor. Migrate to Virtual Machines moves VMs, Database Migration Service moves MySQL, PostgreSQL and SQL Server databases with minimal downtime, and Storage Transfer Service or Transfer Appliance moves bulk data depending on bandwidth. Plan hybrid connectivity before the first wave, test each wave, and check whether licenses such as Windows Server or SQL Server can be brought along or must be pay-as-you-go.
How it shows up
Expect questions on which tool moves which asset, the order of waves, and how to cut over with little downtime.
Common confusion
Transfer Appliance suits very large data sets on limited bandwidth; it is not a database replication tool.
1.5 Envisioning future solution improvementsCloud and technology improvements, changing business needs, cloud-first design.
What it means
This objective covers designing for change. It includes anticipating improvements in cloud and technology, adapting to evolving business needs, and following a cloud-first design approach.
What to understand
A design should leave room to adopt new managed services, add regions or introduce AI features without a rewrite. Loose coupling through APIs and events, infrastructure as code and managed services make later changes cheaper. Cloud-first means new workloads default to managed, elastic services, and legacy systems are modernized in stages rather than copied unchanged forever.
How it shows up
Questions ask which design choice keeps future options open, for example decoupling a monolith or picking a managed database that scales later.
Common confusion
Cloud-first does not mean moving every system immediately; staged modernization is still cloud-first.

High-yield concepts

The terms, rules and formulas this domain turns into questions most often.

  • Google Cloud Well-Architected Framework

    Google's guidance organized into pillars such as operational excellence, security, reliability, cost optimization and performance.

    Remember
    The framework is named directly in objectives 1.2 and 6.1.
    Tested as
    Use it to justify why one design is better balanced than another.
    Trap
    Treating it as a product to deploy rather than a set of design principles.

  • RTO and RPO

    Recovery time objective is how long recovery may take; recovery point objective is how much data loss is acceptable.

    Remember
    Low RPO needs replication; low RTO needs warm or hot standby.
    Tested as
    Disaster recovery scenarios quote both numbers.
    Trap
    Choosing nightly backups when the RPO is minutes.

  • Workload disposition

    Deciding per application whether to build, buy, modify or deprecate it.

    Remember
    Retiring or replacing with SaaS is a valid architecture decision.
    Tested as
    Objective 1.1 lists it explicitly.
    Trap
    Assuming every application must be migrated.

  • Regional vs zonal resources

    Regional managed instance groups, regional GKE clusters and Cloud SQL HA spread across zones.

    Remember
    Zonal resources share the fate of a single zone.
    Tested as
    Availability targets usually point to regional designs.
    Trap
    Using a single-zone VM for a 99.99% service.

  • Spanner

    Fully managed relational database with horizontal scale and strong consistency, including multi-region configurations.

    Remember
    Pick it for global transactional workloads that outgrow Cloud SQL.
    Tested as
    Common answer for worldwide inventory or order systems.
    Trap
    Choosing it for small regional apps where Cloud SQL is cheaper.

  • Cloud Run vs GKE

    Cloud Run runs stateless containers with no cluster to manage; GKE gives full Kubernetes control.

    Remember
    Choose Cloud Run unless you need Kubernetes features or fine-grained node control.
    Tested as
    Compute mapping in 1.3.
    Trap
    Picking GKE for a simple HTTP service with spiky traffic.

  • Spot VMs

    Discounted Compute Engine capacity that Google can preempt at any time.

    Remember
    Suits fault-tolerant batch, rendering and some training jobs.
    Tested as
    Cost questions with interruptible work.
    Trap
    Running a stateful database primary on Spot VMs.

  • Google Cloud Migration Center

    Hub for discovery, assessment, TCO estimates and migration planning.

    Remember
    Assessment comes before choosing migration waves.
    Tested as
    Named in objective 1.4.
    Trap
    Jumping to a migration tool before inventory and dependency mapping.

  • Database Migration Service

    Managed service for minimal-downtime migration of MySQL, PostgreSQL, SQL Server and Oracle sources into Cloud SQL or AlloyDB.

    Remember
    Uses continuous replication so cutover is short.
    Tested as
    Database move questions in migration plans.
    Trap
    Using export and import when the stem demands near-zero downtime.

  • AI Hypercomputer

    Google's integrated GPU and TPU infrastructure, software and consumption models for large AI workloads.

    Remember
    It appears in 1.3 as a design option and in 2.4 for provisioning.
    Tested as
    Pick it when the case needs large-scale training or serving.
    Trap
    Proposing custom training when a hosted Gemini model would meet the need.

  • Model Garden

    Catalog of Google, partner and open models available on Gemini Enterprise Agent Platform.

    Remember
    It lets you choose a model before building anything.
    Tested as
    Model selection in AI design scenarios.
    Trap
    Assuming only Gemini models are available.

  • Case studies

    Two of four published case studies appear in each exam and make up 20 to 30% of questions.

    Remember
    Current set: Altostrat Media, Cymbal Retail, EHR Healthcare, KnightMotives Automotive.
    Tested as
    Read them before exam day.
    Trap
    Studying older case studies such as Mountkirk Games or TerramEarth as if they were current.

Side by side

Concepts candidates mix up, compared on the point that decides the answer.

Migration approaches

ConceptMeaningKey differenceCommon trap
Rehost (lift and shift) Move VMs largely unchanged. Fastest, least benefit from managed services. Picking it when the stem wants lower administration.
Replatform (lift and optimize) Move with small changes, such as a managed database. Some modernization without a rewrite. Confusing it with a full refactor.
Refactor Re-architect to cloud-native services. Most benefit, most time and risk. Choosing it under a hard deadline.
Retire or replace Shut down or swap for SaaS. Removes the workload rather than moving it. Forgetting it is a valid disposition.

Operational databases

ConceptMeaningKey differenceCommon trap
Cloud SQL Managed MySQL, PostgreSQL and SQL Server. Regional, vertical scale, simple lift of existing engines. Expecting global writes.
AlloyDB for PostgreSQL PostgreSQL-compatible database tuned for high performance and analytics. Faster PostgreSQL at higher scale than Cloud SQL. Using it for MySQL or SQL Server workloads.
Spanner Horizontally scalable relational database. Global strong consistency. Paying for it when one region suffices.
Firestore Serverless document database. Flexible schema, mobile and web sync. Using it for complex relational joins.

Compute platforms

ConceptMeaningKey differenceCommon trap
Compute Engine Virtual machines. Full OS control, any software. Choosing it when the stem wants no server management.
GKE Managed Kubernetes. Orchestration for many containers and hybrid fleets. Overkill for one stateless service.
Cloud Run Serverless containers. Scales to zero, no cluster. Assuming it suits long stateful jobs.
Cloud Run functions Event-driven functions. Smallest unit, triggered by events. Using it for a full web application.

Exam traps

Each trap is an answer that looks right. Learn the clue that should change your mind.

  1. Global database by default

    What candidates pick
    Spanner for every relational workload.
    Why it looks right
    It is the most scalable option.
    Why it's wrong
    Spanner costs more and adds design work; a regional application with modest load is better served by Cloud SQL or AlloyDB.

    Clue to watch for Look for words like global, worldwide writes or horizontal scale.

  2. Rewrite before migrating

    What candidates pick
    Refactor everything into microservices first.
    Why it looks right
    It sounds modern and future-proof.
    Why it's wrong
    A fixed deadline such as a lease expiry makes a full rewrite too slow; rehost or replatform first, then modernize.

    Clue to watch for A date or contract deadline in the stem.

  3. Zonal design for high availability

    What candidates pick
    A single managed instance group in one zone with autoscaling.
    Why it looks right
    Autoscaling feels like resilience.
    Why it's wrong
    Autoscaling adds capacity but does not survive a zone outage; a regional managed instance group does.

    Clue to watch for An availability target of 99.9% or higher.

  4. Custom model training

    What candidates pick
    Train a new model on AI Hypercomputer.
    Why it looks right
    It gives maximum control.
    Why it's wrong
    When a hosted Gemini model or a Model Garden model meets the need, training adds cost, time and specialist staff.

    Clue to watch for The stem stresses speed to market or a small AI team.

  5. Ignoring licensing

    What candidates pick
    Move Windows and SQL Server VMs without license review.
    Why it looks right
    Licensing feels like a procurement issue.
    Why it's wrong
    Objective 1.4 lists license and financial impact; bring-your-own-license rules can change host choice and cost.

    Clue to watch for Mentions of existing Microsoft or Oracle agreements.

  6. Cheapest option wins

    What candidates pick
    The lowest-cost storage or compute class.
    Why it looks right
    Cost optimization is a listed requirement.
    Why it's wrong
    Cost matters only after availability, latency and compliance requirements are met.

    Clue to watch for Several requirements listed together with cost last.

Answer framework

Use this sequence for design and case study questions in Section 1.

  1. List every requirement in the stem or case study and mark which are business goals, such as lower cost, and which are technical constraints, such as latency.
  2. Eliminate options that break a hard constraint such as availability target, data residency, latency or a migration deadline.
  3. Among the remaining designs, prefer managed services that reduce administration unless the stem demands low-level control.
  4. Check data and compute fit: access pattern, consistency model, expected scale and the skills of the operating team should all match the chosen service.
  5. Confirm the plan covers the migration path, licensing, cost and room for future change, then choose the simplest option that satisfies every listed requirement at once.

Practice question

An online retailer runs a MySQL-backed catalog on premises and must move to Google Cloud within three months with under ten minutes of downtime. The team wants less database administration but has no time to change application code. What should the architect recommend?

Want more like this? The GCP PCA practice test has questions for every domain.

Checklist

0 of 12

Tick each one you can do without notes. Your ticks stay in this browser.

Study strategy

Begin with the four current case studies in Google's exam guide. For Altostrat Media, Cymbal Retail, EHR Healthcare and KnightMotives Automotive, write one page each listing business requirements, technical requirements and the existing environment, then sketch a target architecture. Doing this before learning individual services gives you a reason to remember each one. Next, build a decision table that maps requirement types such as global consistency, scale to zero or shared file access to the matching storage and compute services.

Spend a full session on availability and disaster recovery: draw zonal, regional and multi-region versions of one application and label the RTO and RPO each achieves. Then study migration by walking through Migration Center assessment, wave planning and the transfer tools for VMs, databases and bulk data. Read the Well-Architected Framework pillar summaries once, since Section 6 returns to operational excellence. Finish with timed scenario practice, explaining aloud why each wrong option fails a specific requirement. Plan for 20 to 25 hours, the largest share of your preparation.

FAQ

How much of the Professional Cloud Architect exam is Designing and planning a cloud solution architecture?

Google lists this section at about 25% of the exam, the largest of six sections. On a 50 to 60 question exam that suggests roughly 13 to 15 questions, though Google does not publish per-section counts.

Which case studies are current?

The standard exam guide lists four: Altostrat Media, Cymbal Retail, EHR Healthcare and KnightMotives Automotive. Each exam uses two of them, and Google says case questions make up 20 to 30% of the exam.

How much AI knowledge does Section 1 expect?

You need to know which AI option fits a requirement: a hosted Gemini model, a Model Garden model, an agent built with Agent Builder, or large-scale training on AI Hypercomputer. Model math is not tested.

Is the Well-Architected Framework tested directly?

Yes. Objective 1.2 names the Google Cloud Well-Architected Framework, and Section 6 tests its operational excellence pillar. Know the pillars and use them to compare designs.

What is Gemini Cloud Assist and why is it in the guide?

Gemini Cloud Assist is an AI assistant in the Google Cloud console that helps design, explain, optimize and troubleshoot resources. The guide lists it under technical requirements and under advising teams.

Is there a passing score?

Google does not publish a passing score for the Professional Cloud Architect exam. Results are reported as pass or fail.

Is this exam being retired or replaced?

No retirement was announced as of October 2026. The certification page notes the exam was updated for recent branding changes and tells candidates to check product names in the current guide.

Other Cloud Computing certifications

Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.

Test your Domain 1 knowledge

Practice case study and design scenarios that ask you to balance availability, cost, migration deadlines and service fit across Google Cloud.

Take the GCP PCA practice test Check your GCP PCA readiness

Keep going