GCP Core Concepts

Before diving in, it’s essential to develop a clear intuition for what the Google Cloud Platform (GCP) is, its core components, and how it organizes and manages resources. This section also introduces the core ideas that will guide how we use GCP throughout this tutorial. This foundation will help you navigate GCP confidently and avoid common pitfalls.

What Is the Google Cloud Platform (GCP)?¶

GCP is a suite of cloud computing^[1] services, including computing, storage, networking, data analytics, machine learning, etc., that runs on the same infrastructure Google uses for products like Search, Gmail, and YouTube. It allows individuals and organizations/enterprises to use Google’s computing infrastructure without owning or managing physical hardware.

At a high level, GCP provides a way to create, configure, and manage computing resources using software in an on-demand manner. Machines, storage, and networks are no longer physical objects, but configurable resources that can be created, modified, and removed as needed.

Why Should You Care About GCP?¶

For AI/ML researchers, GCP provides:

• Compute resources: Virtual machines with GPUs and TPUs.

• Storage: Persistent disks for storing datasets, model artifacts, etc.

• Networking: High-speed connections between resources.

• Flexible pricing: Pay-per-second billing.

• Identity and access management: Fine-grained control over who can create, view, or modify resources.

What is a Resource in GCP?¶

In GCP, a resource is any manageable entity that you can create, configure, and pay fore.

A resource represents something concrete, such as:

• A virtual machine

• A disk or storage bucket

• A network or firewall rule

• A GPU or TPU attached to a virtual machine

• A managed database service instance

Resources have a lifecycle: they can be created, modified, stopped, deleted, and monitored. Creating resources in GCP incurs cost.

GCP Resource Hierarchy¶

GCP organizes resources hierarchically. This structure is crucial for managing access, billing, and organization.

Organization (e.g. *aims.ac.za*)
└── Folders (optional groupings)
        └── Projects (your main workspace)
            └── Resources (virtual machines, storage, etc.)

Organization¶

The Organization sits at the top of the GCP resource hierarchy, representing an institution or company.

Folders (Optional)¶

Folders group projects together. An organization (e.g., AIMS) may use folders to organize projects by cohort, program, or research group.

You typically won’t create folders yourself - focus on projects.

Projects¶

Projects are the core organizational unit. Everything you create lives inside a project.

Key Characteristics:

• Each project has a unique Project ID (immutable, globally unique)

  • Example: trump-project

• Each project has a Project Name (can be changed)

• Each project has a Project Number (assigned by Google Cloud)

• All resources belong to exactly one project

• Billing is tracked per project

• Access control (IAM) is managed at the project level

Core GCP Concepts¶

1. Billing Accounts¶

A billing account is linked to a payment method and pays for resource usage across projects.

2. Quotas¶

Quotas limit resource usage in a project. Google Cloud implements quotas to:

• Prevent accidental overspending

• Ensure fair resource allocation/distribution

• Protect Google Cloud against abuse

Common quotas:

• GPU quotas: Number of GPUs per region

• CPU quotas: Number of vCPUs

• Persistent disk: Total disk space

• API rate limits: Requests per minute

3. Regions and Zones¶

GCP resources are deployed in specific geographic locations.

Region: A specific, independent, and secure geographic location used to deploy resources. Examples include us-central1 (Iowa, USA), europe-west4 (Netherlands), asia-southeast1 (Singapore).

Zones: A deployment area within a region. A region is a collection of zones. Zones have high-bandwidth, low-latency network connections to other zones in the same region. For example, the us-central1 region can have zones like us-central1-a, us-central1-b, us-central1-c.

4. APIs and Services¶

GCP functionality is accessed through APIs. Before using a service, you need to enable its API for your project.

Common APIs you may need to enable:

• Compute Engine API: To create virtual machines

• Vertex AI API: To use managed ML services

• Cloud Storage API: To use object storage services

• Cloud Build API: To use container building services.

5. Identity and Access Management (IAM)¶

IAM controls who can do what in GCP.

Key concepts:

• Principal: Who (user, service account)

  • Example: you@aims.ac.za

• Role: What permissions

  • Owner: Full control

  • Editor: Can modify resources

  • Viewer: Read-only access

• Resource: Where (project, VM, bucket)

Best practices:

• Use least privilege (use only the minimum permissions needed)

• Don’t share credentials

• Use service accounts for automated tasks

🔑 Key Takeaways¶

• ✅ Projects are your main workspace

• ✅ Quotas can block you even with permissions - check and request quota increases early

• ✅ Regions matter for latency, availability, and cost

• ✅ APIs must be enabled before use

• ✅ AIMS manages billing - you focus on research, not payment setup.

🚀 What’s Next¶

Now that you understand GCP’s core concepts, let’s explore the console interface in detail.

📚 References & Further Reading¶

1. Google Cloud Resource Hierarchy

2. Understanding Projects

3. Cloud Quotas

4. Regions and Zones

5. IAM Overview