Google Cloud Architect Course Notes

Professional Cloud Architect design questions begin with requirements, not products. Translate stakeholder goals into availability, latency, compliance, data residency, cost, and supportability targets before selecting compute, networking, data persistence, or recovery patterns.

• Choosing a Google Cloud service before translating business requirements into availability, latency, cost, data residency, and recovery targets.

• If the scenario names downtime tolerance, map it to RTO; if it names acceptable data loss, map it to RPO.

Advanced design patterns test whether the architect can choose the right platform shape for the workload. Compute Engine gives the most VM-level control, GKE fits portable container orchestration, Cloud Run fits stateless containers with request-driven scaling, and Cloud Functions fits small event-driven functions.

• Picking serverless for every workload without checking state, runtime, networking, portability, or control requirements.

• If VM-level control or lift-and-shift compatibility is required, Compute Engine is usually the starting point.

Infrastructure provisioning tests how services are assembled into a secure, scalable platform. Shared VPC centralizes network control in a host project while service projects own workloads, which helps large organizations separate network administration from application delivery.

• Using Private Service Connect as if it replaced IAM or firewall policy; it changes private connectivity, not identity authorization.

• If multiple service projects need centralized networking, choose Shared VPC.

Security architecture starts with least privilege at the right resource level. Grant predefined roles when they fit, create custom roles only for narrow gaps, and place policies where inheritance matches the organization, folder, project, or resource boundary.

• Granting broad predefined roles when a custom role or narrower scope would satisfy least privilege.

• If the clue is policy inheritance across business units, think organization, folders, projects, and resource hierarchy.

Optimization questions ask architects to connect technical signals with business goals. A fast system that wastes budget and a cheap system that violates availability targets are both poor designs, so recommendations should weigh performance, cost, reliability, and stakeholder priorities together.

• Optimizing cost in isolation and accidentally violating reliability or performance targets.

• If the scenario asks for governed self-service deployments, choose Service Catalog.

Implementation management turns an approved architecture into controlled change. The architect should choose delivery patterns that reduce release risk, preserve rollback options, and make each deployment observable before it reaches all users.

• Using ad hoc CLI or Cloud Shell commands as the production release method instead of repeatable pipelines and approvals.

• If the question mentions targets, rollouts, promotion, approval, or rollback, choose Cloud Deploy.

Operations excellence starts before launch. Production services need runbooks, owners, support paths, dashboards, alert policies, log retention, rollback procedures, and clear escalation rules so incidents do not depend on memory or heroics.

• Creating dashboards without actionable alerts, runbooks, owners, or escalation paths.

• If the scenario mentions reliability targets, identify SLIs and SLOs before choosing alerts.