What's changed: Created Professional Cloud Architect Chapter 4 (Domain 2 "Manage and provision" part 2: compute/container config = machine types/MIG/templates, GKE Autopilot/Standard/node pools, Cloud Run revisions/traffic splitting, Spot VMs/CUDs/SUDs; data config and ops/monitoring = Cloud SQL HA/read replicas/Spanner, Cloud Storage lifecycle/uniform access, Cloud Monitoring/Logging/Trace, SLOs/error budgets).
4.1Configuring compute and containers
Understand Compute Engine machine types/MIGs/instance templates, GKE Autopilot/Standard with node pools/workloads, Cloud Run revisions and traffic splitting, and cost-efficient configuration with Spot VMs and committed-use discounts.
Configure the chosen compute correctly for the requirements, balancing performance, scale, and cost via each service feature.
4.1.1Compute Engine and GKE
In Compute Engine, pick a fitting machine type (general/compute-optimized/memory-optimized) and scale/self-heal identical VMs in a MIG from an instance template. In GKE, choose Autopilot (Google manages nodes) or Standard (you configure nodes), mix machine configs via node pools, and run workloads (Deployments/Services). Map "let Google run nodes too = GKE Autopilot" and "fine node control = GKE Standard."
4.1.2Cloud Run and cost efficiency
In Cloud Run, each deploy creates a revision, and traffic splitting enables gradual rollout (canary) and rollback. For cost efficiency, combine Spot VMs (deep discount) for interruptible batch, committed-use discounts (CUDs) for steady workloads, and the automatic sustained-use discounts (SUDs) for long uninterrupted use. Map "interruptible = deep discount Spot VMs" and "commit 1–3 years = CUDs."
Common: requirement → config. E.g., "let Google manage nodes" = GKE Autopilot; "gradual canary release" = Cloud Run traffic splitting; "cheap interruptible batch" = Spot VMs; "discount steady usage" = committed-use discounts (CUDs); "scale/self-heal identical VMs" = MIG + instance template.
Watch the mix-ups: (1) Do not confuse GKE Autopilot (managed nodes) and Standard (you configure nodes). (2) Spot VMs can be preempted—unfit for stateful or always-required workloads. (3) CUD (committed discount) and SUD (automatic discount) differ.
The fully managed PaaS App Engine offers a standard environment (Google-managed language runtimes, can scale to zero) and a flexible environment (any container), with traffic splitting for easy canary/A-B releases—an option for web/API backends that minimize infrastructure management.
4.1.3Section summary
- Compute Engine = machine type + instance template + MIG for scaling/self-healing
- GKE = Autopilot (managed nodes)/Standard (self-configured) + node pools
- Cloud Run = revisions + traffic splitting; optimize cost via Spot VMs/CUDs/SUDs
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. To let Google manage even the nodes in GKE, which mode best fits?
Q2. In Cloud Run, to release a new version gradually and roll back on issues, what do you use?
Q3. To run interruptible, re-runnable batch as cheaply as possible, which is best?
Q4. To cut cost for steady workloads by committing 1–3 years, which is best?
Q5. To scale identical VMs with demand and self-heal on failure, what do you use?
Q6. Which is a correct caveat when using Spot VMs?
Keep track of your progress
The full study guide is free to read. Sign up free to practice with the question bank, track what you have read, review your mistakes, and highlight passages.