Skip to content

Performance & data

Google Cloud

Google Cloud Development

We build and run the backend half of an ecommerce operation on Google Cloud — serverless containers on Cloud Run, images built by Cloud Build, credentials in Secret Manager, and every service holding an identity of its own.

A commit becoming a revision, and the revision it can go back toPRF · 05
Overview

Serverless containers, not a pile of VMs

Most of what an ecommerce business needs from a cloud is a handful of small services that answer HTTP, hold some state, talk to Shopify and to each other, and cost nothing when nobody is shopping. Cloud Run does exactly that, and it does it without asking you to run a cluster.

What separates a Cloud Run project that ages well from one that becomes somebody's problem is the boring layer: a reproducible build, an artifact you can point at, secrets that are not in the repo, a service account per service rather than one owner key, and a deploy you can undo. That is the layer we build.

Scales to zero
A service that only runs when it is called costs nothing overnight. For catalogue jobs and internal tools, that is most of the day.
Identity per service
Each service runs as its own IAM service account with only the permissions its job needs, so one compromised token is not a key to everything.
A build you can trace
Cloud Build produces a tagged image in Artifact Registry, so every running revision maps back to a commit rather than to somebody's laptop.
What's included

What the engagement covers

  • Cloud Run services with containers, revisions and traffic control
  • Cloud Build pipelines from a GitHub repository
  • Artifact Registry for versioned, traceable images
  • Secret Manager for credentials, wired into the runtime
  • IAM roles and per-service runtime service accounts
  • Cloud Storage buckets with uniform access, CORS and lifecycle rules

Technologies

  • Google Cloud Run
  • Cloud Build
  • Artifact Registry
  • Secret Manager
  • IAM
  • Cloud Storage
  • Node.js
  • Docker
Use cases

Where this helps most

  • An API behind a storefront

    A Node.js service your Shopify theme or app calls, that needs to be fast, cheap at idle and safe to deploy on a Friday.

  • Catalogue and file workloads

    Imports, exports, image processing and document uploads that belong in object storage and a job, not in a web request.

  • Shared services across several stores

    One backend serving more than one storefront, where per-store configuration and per-service identity both matter.

Process

How we deliver

  1. Shape the services

    We decide what is one service and what is three, what needs state, and what can be a job rather than a permanently running process.

  2. Build and registry

    A Dockerfile, a Cloud Build trigger and an Artifact Registry repository, so a push becomes an image and an image becomes a revision.

  3. Identity and secrets

    Each service gets a service account and reads its configuration from Secret Manager. Nothing sensitive lives in the repository or in a deploy script.

  4. Run it

    Health checks, structured logs, revision-based rollback, and a written note of what each service is and who it talks to.

FAQ

Google Cloud Development

Do we need Kubernetes?
Almost certainly not. Cloud Run gives you containers, autoscaling and a URL without a cluster to operate. We would only reach for GKE if you had a workload Cloud Run genuinely cannot host, and we would say so rather than default to it.
Can Cloud Run talk to our database?
Yes — over the network with the right IAM and networking configuration, whether the database is Cloud SQL, a managed provider elsewhere, or something you host. Which option is right depends on latency and on where your data is allowed to live, so it is a decision we make with you rather than for you.
What does this cost to run?
That depends entirely on your traffic shape, and we will not quote a figure we cannot stand behind. What we can say is what drives it: request count, allocated CPU and memory, and whether the service is allowed to scale to zero. We set those deliberately and show you where they land.
Do you also migrate existing apps?
Yes. If the app is currently on Heroku, that has its own page with its own plan — inventory, containerisation, a parallel run and a reversible cutover.
Let's build

Ready to start with google cloud development?

Tell us about your store and your goals. We'll come back with a clear, honest plan and a transparent quote.