A team building a simple internal tool deploys it on a full Compute Engine VM they now have to patch and monitor forever, when Cloud Run would have handled the exact same workload with zero server management and automatic scale-to-zero included. This pillar covers Google Cloud's full compute spectrum - from full-control virtual machines to fully managed serverless platforms - so the compute model is chosen based on the actual workload, not habit.
What This Pillar Covers
- Choosing between Compute Engine, GKE, Cloud Run, and Cloud Functions for a given workload
- Configuring Managed Instance Groups with autoscaling based on real load
- Deploying to GKE in Autopilot versus Standard mode
- Deploying serverless containers with Cloud Run
- Reducing compute costs with Spot VMs and custom machine types
- Managing container images with Artifact Registry
Who This Is For
Cloud administrators, DevOps engineers, and backend developers responsible for deploying applications, choosing the right compute service for a workload's actual traffic pattern, and keeping compute costs proportional to real usage.
Why This Matters in Production
The compute decision - Compute Engine versus GKE versus Cloud Run versus Cloud Functions - is the single most heavily weighted judgment call on the Associate Cloud Engineer exam, and the first real architecture decision most GCP newcomers face. Reaching for GKE by default for a single simple container adds ongoing cluster management overhead a workload that size never actually needed.
Prerequisites
- Completion of GCP Fundamentals and Resource Governance, or equivalent familiarity with Projects and IAM
- Basic understanding of Linux server administration
- Familiarity with containers and Docker is helpful for the GKE and Cloud Run topics