Cloud computing in simple terms means using someone else’s hardware to build our projects.
It’s extremely resourceful for projects that expect drastic changes in compute requirements for short bursts because cloud enables us to pay as we go for compute resources.
Cloud is of two types: private cloud and public cloud.
Private cloud is when a compute cloud is built only for your firm or project, often on-premises. It’s commonly used in banks.
Public cloud is a cloud service provided by vendors for everyone — examples are GCP, AWS, Azure, IBM Cloud.
Sometimes, a hybrid is used to avoid vendor lockouts.
Over time, most projects have turned into cloud-native ones because as requirements grew drastically, procuring large sets of hardware became impossible, so companies started building directly for the cloud.
Google Cloud Offerings
GCP offers three main kinds of services:
- IaaS
- PaaS
- SaaS
They follow this hierarchy:
IaaS (bottom) → PaaS (mid) → SaaS (client offering)
Cloud supports all kinds of architectures such as monoliths, modular monoliths, or services.
GCP Security Model
GCP security has six layers:
- Hardware infrastructure layer
- Service deployment layer
- User identity layer
- Storage service layer
- Internet communication layer
- Operations layer
Global Coverage
GCP is offered across seven locations worldwide, spanning multiple continents, and follows a location → region → zone pattern for denotation.
Resource Hierarchy
Cloud follows a hierarchy structure:
- Level 0: Resources
- Level 1: Projects
- Level 2: Folders
- Level 3: Organization node
Policies are applied at levels 1–3, usually set at level 3 and inherited downward.
IAM Policy
IAM basically defines who is who and who does what in a project.
Roles are assigned to users or engineering teams and come in three forms:
- Basic
- Predefined
- Custom
Virtual Private Cloud (VPC)
VPC (Virtual Private Cloud) is basically your own cloud inside a public vendor with your own rules.
Each VPC has subnets, which can be considered as smaller parts of a big VPC — like small rooms inside one building.
VPCs in GCP allow firewall-free internal access between subnets (unlike AWS which doesn’t allow this by default).
Compute Engine
Compute Engine is where you create and run Virtual Machines (VMs) on Google’s infrastructure.
Each VM is a full-fledged OS, Linux or Windows, created through console, CLI, or API.
Google Cloud Marketplace provides pre-built images for fast deployment.
Billing is per second with automatic discounts for long-running workloads.
For big datasets or batch jobs, Preemptible and Spot VMs give heavy discounts.
Custom machine types let you pay for exactly what you use.
Load Balancing
A load balancer is like a chef managing a busy kitchen — it distributes the user traffic among multiple servers (the cooks).
If more customers arrive, the chef adds more cooks.
Google Cloud Load Balancers are software-defined, global, and automatically manage failovers and spikes without prewarming.
Two main types:
- Application Load Balancers – HTTP/HTTPS (Layer 7)
- Network Load Balancers – TCP/UDP (Layer 4)
Cloud DNS and CDN
DNS is the internet’s phonebook — name to IP.
Cloud DNS is Google’s managed, low-latency, globally replicated system.
Cloud CDN uses edge caching to serve content faster from locations closest to users, saving cost and reducing VM load.
Connecting Networks
When organizations grow, they link multiple clouds or regions together.
Main ways:
- Cloud VPN – creates encrypted tunnels between clouds.
- Direct and Carrier Peering – physical or ISP-based connections to Google’s network.
- Dedicated/Partner Interconnect – private links with high bandwidth and 99.99% SLA.
- Cross Cloud Interconnect – direct high-speed connection between GCP and other clouds.
Google Cloud Storage Options
Different workloads, different storage:
- Cloud Storage – for unstructured data (images, backups, media).
- Cloud SQL – managed relational databases (MySQL, PostgreSQL, SQL Server).
- Spanner – globally scalable relational database built for high consistency.
- Firestore – NoSQL for mobile and web apps (like a smarter MongoDB).
- Bigtable – wide-column NoSQL for analytics and IoT-scale data.
Containers in the Cloud
Containers solve the heavy lifting problem of VMs — smaller, faster, isolated environments for code.
Analogy:
- A container is a flight journey.
- The Pod is the airplane running the flight.
- The Node is the airport hosting multiple airplanes.
- The Control Plane is the air traffic control keeping order.
- The Cluster is the whole airline network.
- And the Service is the airline brand — users don’t care which plane they’re on, they just get where they need to.
Kubernetes and GKE
Kubernetes manages these containers automatically — scaling, healing, and balancing them.
GKE (Google Kubernetes Engine) is Google’s managed Kubernetes service that handles the control plane for you.
Two modes:
- Autopilot – Google manages configuration and scaling.
- Standard – full user control for customization.
Cloud Run and Knative
Cloud Run runs stateless containers without servers — just deploy your container and Google handles the rest.
Built on Knative, it scales to zero when idle and auto-expands on demand.
You can run web apps, APIs, or background tasks — in any language that runs on Linux 64-bit.
Cloud Run Functions are lightweight event-based functions that trigger from HTTP requests or Cloud events, billed per 100 milliseconds.
(My personal favorite as a starting point for MVP, just drop your Docker image of backend for scaling on CRF, and it works amazingly)
Top comments (0)