gcp

name	gcp
description	Google Cloud Platform with Cloud Run, GKE, and BigQuery. Use for GCP infrastructure.

Google Cloud Platform (GCP)

GCP is known for its data analytics (BigQuery) and being the home of Kubernetes. 2025 highlights Vertex AI for rapid model deployment and Cloud Run for serverless everywhere.

When to Use

• Big Data: BigQuery is arguably the best data warehouse for speed and ease of use.
• Kubernetes: GKE (Google Kubernetes Engine) often leads in features and stability (Autopilot).
• AI: TPU access and Vertex AI for training large models.

Core Concepts

Projects

The primary unit of isolation. Resources belong to a Project. Projects belong to Folders/Organization.

Global VPC

Unlike AWS/Azure, GCP VPCs are global. Subnets are regional. A single VPC can span the world.

IAM

Permissions are granted to Members (Users/Service Accounts) on Resources via Roles.

Best Practices (2025)

Do:

• Use Cloud Run: The default computed choice for stateless containers. Scales to zero, fast startup.
• Use Workload Identity: Let GKE workloads impersonate Service Accounts securely.
• Shared VPC: For orgs, use a Shared VPC in a Host Project to centralize networking.

Don't:

• Don't use owner role: It's too broad. Use precise roles (storage.objectViewer).

References

• Google Cloud Documentation