Google Cloud Fundamentals: Learning Notes Summary (Part 2 of 4)

Course: Google Cloud Fundamentals: Core Infrastructure | Google Cloud Skills Boost https://www.cloudskillsboost.google Focus: Virtual Machines and Networks in the Cloud

Introduction

This is the second part of my systematic GCP learning journey. While this article stands alone, I recommend reading Part 1 for foundational concepts like resource hierarchy and IAM.

Part 2 focuses on compute and networking—the building blocks of cloud infrastructure.

What’s Covered

This article explores GCP’s compute and networking capabilities:

• Virtual Private Cloud (VPC) fundamentals
• Compute Engine as IaaS
• VM types and pricing models
• Autoscaling strategies
• VPC networking features
• VPC connectivity (peering, shared VPC)
• Load balancing options
• DNS and edge caching
• Connecting VPCs to external networks

Virtual Private Cloud (VPC): Your Private Space in the Public Cloud

A Virtual Private Cloud (VPC) is your own isolated network environment within Google Cloud. Think of it as having a private cloud with all its benefits—security, control, customization—but hosted on Google’s public infrastructure.

What Makes Google Cloud VPC Special

Global Scope: Unlike traditional networks, Google Cloud VPCs span the entire globe by default.

Regional Subnets: You can segment your VPC into subnets, each located in any Google Cloud region worldwide.

Cross-Zone Flexibility: Subnets can extend across multiple zones within a region, meaning resources in different zones can belong to the same subnet.

VPC (Global)
├── Subnet: us-central1 (Regional)
│   ├── Zone: us-central1-a (Resources here)
│   └── Zone: us-central1-b (Resources here)
└── Subnet: europe-west1 (Regional)
    ├── Zone: europe-west1-a
    └── Zone: europe-west1-b

This architecture simplifies creating global applications without complex network configurations.

Compute Engine: Google’s IaaS Solution

Compute Engine lets you create and run virtual machines on Google’s infrastructure with the performance and functionality of physical servers.

Key Features

• No upfront investment: Pay only for what you use
• Massive scale: Run thousands of virtual CPUs
• Full OS control: Configure CPU, memory, storage, and operating system
• Consistent performance: Built on Google’s reliable infrastructure

Creating Virtual Machines

You can create VMs through:

• Google Cloud Console (web interface)
• Google Cloud CLI (command line)
• Compute Engine API (programmatic access)

Operating System Options

• Linux and Windows Server images from Google
• Customized versions of provided images
• Your own operating system images

Quick Start Tip: Use the Cloud Marketplace for pre-configured solutions from Google and third-party vendors. Note that third-party images may include additional licensing costs for commercial software.

Compute Engine Pricing: Understanding Your Costs

Google’s pricing model is designed for flexibility and cost savings:

Three Types of Compute Engine VMs

Choose the VM type based on your workload characteristics:

Persistent (On-Demand) VMs

Use case: Standard workloads requiring continuous availability

• Run as long as you need
• Full control over termination
• No interruptions
• Standard pricing

Preemptible VMs

Use case: Batch jobs, fault-tolerant workloads, development/testing

• Maximum runtime: 24 hours
• Cost savings: 60-91% discount
• Limitation: Google can stop them at any time
• Ideal for: Jobs that can pause and resume

Example scenarios:

• Data processing pipelines
• Rendering tasks
• Scientific computations
• CI/CD test environments

Spot VMs

Use case: Same as Preemptible, but with more flexibility

• Same pricing as Preemptible VMs (currently)
• No maximum runtime (unless you specify one)
• Stopped only when Google needs capacity elsewhere
• Additional benefits: Automated workload scheduling, cluster autoscaling

Key difference: Spot VMs are the evolved version of Preemptible VMs with enhanced features.

Custom Machine Types

Beyond standard VM configurations, you can create custom machine types by specifying:

• Exact number of virtual CPUs needed
• Precise amount of memory required

Note: Maximum CPUs per VM depend on the machine family and your zone-specific quota.

Autoscaling: Dynamic Resource Management

Autoscaling automatically adjusts your VM count based on demand, ensuring optimal performance and cost efficiency.

How It Works

Low Traffic → Fewer VMs → Lower Costs
    ↓
Traffic Increases → Autoscaling Adds VMs
    ↓
High Traffic → More VMs → Maintained Performance
    ↓
Traffic Decreases → Autoscaling Removes VMs
    ↓
Back to Low Traffic → Cost Optimization

Scaling Strategies

Scale Out (Horizontal): Add more VMs to handle increased load—this is how most Google Cloud customers start.

Scale Up (Vertical): Use larger VMs with more resources—ideal for specific workloads like:

• In-memory databases
• CPU-intensive analytics
• Single-threaded applications

VPC Networking Capabilities: Built-In Intelligence

Google Cloud VPCs come with networking features that you don’t need to provision or manage separately.

Automatic Routing

No router to configure: VPCs include built-in routing tables automatically.

What they do:

• Forward traffic between instances in the same network
• Route across subnetworks
• Enable communication between Google Cloud zones
• Work without external IP addresses

Native Firewall Protection

No firewall appliance needed: VPCs provide a global distributed firewall out of the box.

Control options:

• Restrict incoming traffic (ingress rules)
• Restrict outgoing traffic (egress rules)
• Define rules using network tags on instances
• Apply rules globally or to specific resources

Convenience: Network tags make firewall management simple—tag your instances and apply rules to those tags.

Connecting VPCs: Peering and Sharing

VPC Peering

Problem: VPCs are project-specific. How do you connect resources across projects?

Solution: VPC Peering establishes private connectivity between VPCs.

Project A (VPC-A) ←→ VPC Peering ←→ Project B (VPC-B)

Benefits:

• Private communication between projects
• No internet exposure
• Lower latency
• Reduced egress costs

Shared VPC

Use case: Organizations with multiple projects needing centralized network management.

How it works:

1. Designate one project as the host project
2. Attach other service projects to it
3. Resources in service projects use the host project’s VPC network

Host Project (Shared VPC Network)
├── Service Project 1 (uses host VPC)
├── Service Project 2 (uses host VPC)
└── Service Project 3 (uses host VPC)

Advantages:

• Centralized network administration
• Consistent security policies
• Simplified network architecture
• Internal IP communication across projects

Cloud Load Balancing: Distributing Traffic Intelligently

Load balancing distributes incoming traffic across multiple VM instances, preventing any single instance from becoming overwhelmed.

Why Google’s Load Balancing is Different

Load Balancing Options

Choose based on your traffic type and architecture:

For Internet-Facing Traffic

For Internal Traffic

Example architecture:

Internet Users
    ↓
HTTP(S) Load Balancer (Global)
    ↓
Web Tier (Compute Engine VMs)
    ↓
Internal Load Balancer (Regional)
    ↓
Application Tier (Compute Engine VMs)

Cloud DNS: Managed Domain Name System

Cloud DNS is Google’s managed DNS service running on the same infrastructure that powers Google’s own services.

Key Features

• Low latency: Fast DNS resolution
• High availability: 100% uptime SLA
• Global redundancy: Served from locations worldwide
• Cost-effective: Pay only for what you use
• Programmable: Manage via Console, CLI, or API

Scale

Publish and manage millions of DNS zones and records without worrying about infrastructure.

Cloud CDN: Edge Caching for Performance

Cloud CDN (Content Delivery Network) uses edge caching to store content closer to your users, reducing latency and improving performance.

Benefits

Easy Setup

Once HTTP(S) Load Balancing is configured, enabling Cloud CDN is just one checkbox.

User Request → Nearest Edge Location (cached content) → Fast Response
    ↓ (cache miss)
Origin Server → Edge Cache → User

Connecting to External Networks

Google Cloud offers multiple ways to connect your VPC to on-premises networks or other cloud providers:

Connection Options Compared

Choosing the Right Option

VPN (Virtual Private Network)

• Uses IPsec protocol
• Encrypted tunnel over public internet
• Most flexible and cost-effective
• Good for: Getting started, lower bandwidth needs

Direct Peering

• Router in same datacenter as Google point of presence
• 100+ Google PoPs worldwide
• Work with Carrier Peering partners if needed
• Good for: Direct traffic exchange with Google

Dedicated Interconnect

• One or more direct, private connections
• Can be backed up by VPN for reliability
• 10 or 100 Gbps per connection
• Good for: High bandwidth requirements, sensitive data

Partner Interconnect

• Connect through supported service provider
• For locations without Dedicated Interconnect access
• For workloads not needing full 10 Gbps
• Good for: Smaller bandwidth needs, remote locations

Concluding Thoughts

Google Cloud’s compute and networking services provide a powerful, flexible foundation for building applications. The combination of global VPCs, intelligent load balancing, autoscaling, and multiple connectivity options means you can architect solutions that are both performant and cost-effective.

Part 3 will explore storage options in Google Cloud, covering everything from object storage to databases.

If you found this helpful, please share it with others learning GCP!