Designing and planning a cloud solution architecture · Topic 1 of 8
Business Requirements and Design Trade-offs
Turning business goals into a cloud design: functional and non-functional requirements, cost, KPIs and ROI, build or buy decisions, integration and data movement.
- 4 min
- Read for non-functional requirements: availability, latency, residency, cost, team skills, deadlines.
- Build what differentiates, buy commodities, modify what blocks goals, retire what nobody needs.
- Match the integration pattern to freshness and volume: API, messaging, change data capture or batch.
What it is
Every architecture starts with what the business needs: the use case, the budget, the deadline, the rules it must follow and how success will be measured. The exam gives you these as a scenario and expects you to design for all of them, then pick the trade-off the business would accept.
How it works
Functional and non-functional requirements
| Kind | Question it answers | Examples |
|---|---|---|
| Functional | What must the system do? | Take dealer orders, summarize videos, export claims to insurers |
| Non-functional | How well, how safely, at what cost? | 99.9% availability, under 50 ms reads, EU data residency, monthly budget, no servers to patch |
Most wrong answers on the exam meet the functional need and break a non-functional one. A design that works but needs a team the company does not have, or stores data in the wrong region, is still wrong.
Business constraints that decide the answer
| Constraint in the stem | What it usually points to |
|---|---|
| "Small team", "no operations staff", "lower administration cost" | Managed and serverless services: Cloud Run, GKE Autopilot, Cloud SQL, BigQuery |
| "Lowest cost" with flexible timing | Batch over online, Spot VMs, colder storage classes, committed use discounts for steady load |
| "Cannot change the application" | Rehost or a compatible managed service (Cloud SQL for MySQL, Memorystore) |
| "Must keep a system on premises" | Hybrid connectivity and an integration layer, not a migration |
| "Regulated data" | Region choice, encryption keys, audit logs, compliance controls |
| "Deadline" (lease ends, launch date) | The fastest safe path, often rehost first and modernize later |
Workload disposition: build, buy, modify or deprecate
For each workload, decide what happens to it:
| Option | When it fits |
|---|---|
| Build | The capability differentiates the business and nothing on the market fits |
| Buy (SaaS or Marketplace) | A commodity capability: email, CRM, ticketing |
| Modify (replatform or refactor) | The workload matters, but its current form blocks scale or cost goals |
| Deprecate (retire) | Few users, duplicated by another system, or replaced by a new product |
When a stem says a legacy system "will be replaced over the next several years" and "is not being upgraded", leave it where it is and connect to it. Moving it is wasted effort.
Integration with external systems and data movement
| Pattern | Use it for |
|---|---|
| Synchronous API (Apigee, API Gateway) | Partners or dealers calling functions in real time, with quotas and keys |
| Asynchronous messaging (Pub/Sub) | Decoupling producers and consumers, absorbing spikes, fan-out to many teams |
| Change data capture (Datastream) | Keeping an analytics copy current while the source database stays in place |
| Batch file transfer (Storage Transfer Service, Transfer Appliance) | Large one-off or scheduled moves of files |
Ask how fresh the data must be (seconds, minutes, nightly), how much there is and whether the source must stay online.
Cost optimization as a requirement
Cost appears in almost every case study. Common levers: right-size from real usage, use committed use discounts for steady load, Spot VMs for interruptible work, autoscaling to zero where possible, storage lifecycle for ageing data and batch pricing for work that is not urgent. The cheapest option still has to meet the availability and compliance requirements.
Success measurements
| Measure | Example |
|---|---|
| KPI | Order confirmation time, catalog items onboarded per week, call center minutes |
| ROI | Savings or new revenue against the cost of the project |
| Technical metrics | Availability, latency, error rate, deployment frequency |
A good design states how success will be measured before it is built, so the business can fund the next phase on evidence.
Security, compliance and observability from the start
Business requirements include protecting customer data and proving it. Plan identity, encryption, audit logging and monitoring in the first design, not after go-live.
How the exam asks about this
- "Which design meets the requirements at the lowest cost?" Check every requirement first, then compare cost.
- "The company has three platform engineers." Prefer managed services.
- "Leadership will fund the next phase only after measured savings." Baseline the current process and measure a pilot.
What to remember for the exam
- Read for non-functional requirements: availability, latency, residency, cost, team skills, deadlines.
- Build what differentiates, buy commodities, modify what blocks goals, retire what nobody needs.
- Match the integration pattern to freshness and volume: API, messaging, change data capture or batch.
- Cost is a lever, never permission to break another requirement.
- Define KPIs and ROI before the build.
An option that is technically excellent but needs skills or staff the company does not have is a trap.
Common mistakes
Solving only the functional requirement. The design works but misses residency, availability or budget.
Migrating what should stay. Systems the stem says will be replaced later are connected, not moved.
Choosing the cheapest option blindly. It must still meet the stated service levels.
Check yourself
A company has no operations staff and wants lower administration cost. What kind of service should you favour?
Managed or serverless services such as Cloud Run, GKE Autopilot, Cloud SQL and BigQuery.
Analysts need changes from an on-premises MySQL database in BigQuery within minutes while the database stays in place. Which pattern fits?
Change data capture with Datastream.
What does "deprecate" mean in workload disposition?
Retire the workload because it is no longer needed or is replaced.
Topic 1 of 8 in Designing and planning a cloud solution architecture
Test yourself on Designing and planning a cloud solution architecture
The full Google Cloud Professional Cloud Architect question bank has 877 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Google Cloud. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.