Google Cloud proofs of concept

Between roles in late 2023, I took on a freelance contract building proofs of concept on Google Cloud. Most of my experience until then was on AWS, so this was a great chance to learn a second cloud properly.

What a good proof of concept does

A proof of concept isn't a demo. It's a small, honest answer to a client's real question: will this work for us? So each one had to run real workloads, not slides.

What I built

• Kubernetes on GKE, with workloads and clusters set up the way the client would actually run them.
• Compute Engine for the parts that didn't belong in containers.
• AlloyDB as a managed, PostgreSQL-compatible database, to show what a move off self-managed databases could look like.
• A proper VPC, with networking designed first instead of bolted on later.
• Argo CD for GitOps, so every deployment came from a Git repo and the cluster stayed in sync with it.
• Telemetry, so the client could see how things behaved, not just that they ran.

AWS vs Google Cloud, from someone who uses both

The concepts transfer almost one to one: networks, compute, managed databases, identity. What changes is the naming and the defaults. Once you stop translating word for word and learn each cloud's way of thinking, it gets easy.

What I learned

• Short projects need clear success criteria on day one.
• GitOps makes a proof of concept easy to hand over. The repo is the documentation.
• Learning a new cloud is fastest when you have to deliver something real on it.