Designing and planning a cloud solution architecture · Topic 7 of 8
Future Improvements and Cloud-First Design
How a design evolves: cloud-first principles, modernizing after a migration, adopting managed and AI services, and staying ready for new needs.
- 4 min
- Cloud-first: managed, serverless, automated, designed for failure, paid for by use.
- Rehost to meet a deadline, then modernize in planned steps.
- Decouple with APIs and events so parts can change independently.
What it is
An architect plans beyond go-live. The exam asks how a design should evolve when the business grows, new Google Cloud capabilities appear or a temporary choice (such as a rehosted VM) should be modernized. Cloud-first design means new work starts on managed cloud services by default, and on-premises or self-managed options need a reason.
How it works
Cloud-first principles
| Principle | In practice |
|---|---|
| Managed before self-managed | Cloud SQL before MySQL on VMs; Managed Service for Prometheus before your own Prometheus servers |
| Serverless where it fits | Cloud Run, BigQuery, Pub/Sub, Dataflow scale without capacity planning |
| Automate everything | Infrastructure as code, CI/CD, policy as code |
| Design for failure | Assume zones fail; use regional and multi-region services |
| Pay for use | Autoscaling and scale to zero instead of fixed capacity |
| Data as a product | Shared, governed data in BigQuery instead of silos |
Modernizing after a migration
A deadline often forces a rehost. The future-improvements plan then lists what to modernize next:
| Starting point | Next step |
|---|---|
| VMs running containers by hand | GKE or Cloud Run |
| Self-managed databases on VMs | Cloud SQL, AlloyDB or Spanner |
| Nightly SFTP and batch ETL | Event-driven ingestion with Pub/Sub, Eventarc and Dataflow |
| Email alerts from scattered tools | Cloud Monitoring alerting policies, SLOs and paging integrations |
| Siloed data in many databases | A governed analytics platform in BigQuery |
| Manual content work | Gemini for extraction, summaries and generation, with human review |
Planning for evolving business needs
- Growth: choose services that scale horizontally so growth does not force a redesign.
- New regions or regulations: design with location controls (organization policies, regional resources) so expanding into a new geography is configuration, not a rebuild.
- New channels and partners: put APIs behind Apigee so new partners onboard through a portal.
- New AI features: keep data clean, labeled where possible and accessible in BigQuery or Cloud Storage so models and agents can use it later.
Keeping up with cloud improvements
Google Cloud adds services and renames products often. A healthy architecture:
- isolates components behind APIs and events, so one part can be replaced alone;
- tracks new capabilities through reviews against the Well-Architected Framework;
- uses tools such as Active Assist recommendations and Gemini Cloud Assist to find idle resources, rightsizing and security improvements.
When an answer adds a new self-managed component (your own Kafka, your own Prometheus, your own SIEM) and a managed equivalent exists, the exam usually prefers the managed one.
Balancing change against stability
Not every improvement is worth doing now. Weigh the benefit against the risk, the team's skills and the business calendar. A company in its busiest season defers risky changes; a company with a hard deadline rehosts first and modernizes later.
How the exam asks about this
- "After migration, which change reduces operational effort?" Replace a self-managed component with a managed service.
- "The company plans to expand to new countries." Design location controls and regional deployment patterns now.
- "Leadership wants AI features next year." Consolidate and govern data today.
What to remember for the exam
- Cloud-first: managed, serverless, automated, designed for failure, paid for by use.
- Rehost to meet a deadline, then modernize in planned steps.
- Decouple with APIs and events so parts can change independently.
- Use recommendations and reviews to find the next improvements.
- Every improvement still has to meet cost, risk and compliance needs.
"Modernize everything during the migration" is rarely right when a deadline is close. It adds risk to the move.
Common mistakes
Treating the migration as the end state. A rehosted VM is a starting point.
Adding self-managed tools when managed ones exist.
Ignoring data readiness before promising AI features.
Check yourself
What does cloud-first design mean?
New work starts on managed cloud services by default; self-managed or on-premises options need a reason.
A company rehosted VMs to meet a lease deadline. What is a sensible next step?
Modernize in planned steps, for example moving databases to Cloud SQL and containers to GKE or Cloud Run.
How do APIs and events help future change?
They decouple components so one part can be replaced without changing the others.
Topic 7 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.