Plan and Implement a Cloud Solution
What this domain really tests
Design and deployment decisions: choosing compute (Compute Engine, GKE, Cloud Run, App Engine), storage (Cloud Storage classes, Persistent Disk, Filestore), and databases (Cloud SQL, Spanner, Firestore, Bigtable), plus networking (VPC, subnets, firewall rules, load balancers). Questions reward matching workload characteristics to the right product.
Concepts to master
- Compute choice: VMs for full control; GKE for container orchestration; Cloud Run for stateless containers with scale-to-zero; App Engine for managed app hosting.
- Storage classes: Standard, Nearline (30d), Coldline (90d), Archive (365d) - minimum storage durations drive cost traps.
- Databases: Cloud SQL (regional relational), Spanner (global strongly consistent relational), Firestore (document/NoSQL), Bigtable (wide-column, high throughput).
- Networking: subnets are regional; firewall rules are stateful and apply by tag/service account; global vs regional load balancers.
Traps
- Picking Spanner when Cloud SQL suffices, or Cloud SQL when global strong consistency is required.
- Ignoring minimum storage duration when moving objects to Coldline/Archive - early deletion incurs charges.
- Assuming a firewall rule applies to all instances; it targets tags or service accounts.
- Confusing regional and global load balancers for cross-region traffic.
Rule of thumb
Start from the workload's constraints - consistency, latency, scale, and cost - then pick the managed service that meets them with the least operational burden.
Worked example
A globally distributed, strongly consistent transactional database with horizontal scale → Cloud Spanner. A single-region MySQL app → Cloud SQL.