Production patterns for backend engineers who are tired of hello-world tutorials.
Senior full-stack engineer and engineering lead with 5+ years building cloud-native systems at scale. I write about the things the docs don't cover — GCP, FastAPI, Terraform, event-driven architecture, and AI-powered backends. No hello-world demos.
Markdown supported · ⌘↵ to post
That checklist matches how I keep deciding. If the service is just request-driven, Cloud Run is the honest start. The moment I need a worker sitting on a queue all day, the cluster stops being optional.
The idle-cost comparison is the part most Cloud Run vs GKE writeups skip: a low-traffic service on GKE bills for the node whether requests arrive or not, while Cloud Run bills per request. Starting on Cloud Run and moving only when a real requirement forces it keeps that bill honest.
Mahdi Jabinpour
The workload-first framework is useful. One branch needs an update, though: Cloud Run now supports GPUs, so a GPU requirement alone no longer makes GKE the default answer. Google’s “GPU support for services” documentation covers the available configurations. I would make that decision depend on the accelerator, runtime and operational constraints rather than GPU versus no GPU.