How Cloud Computing Works
IntermediateCloud computing delivers computing resources — servers, storage, databases, networking, software — over the internet on demand, billed by usage rather than owned outright. Virtualisation technology allows one physical server to run dozens of independent virtual machines simultaneously. Providers like AWS, Azure, and GCP operate millions of servers in data centres worldwide, and you rent fractions of them instantly through APIs. Cloud computing enables businesses to scale from zero to global in minutes without buying hardware.
Think of cloud computing like electricity from the grid
Before public electricity grids, factories built their own generators. It was expensive, required expertise, and the generator sat idle during off-peak hours. The electricity grid let companies simply plug in and pay for what they used — the complexity of generation and distribution disappeared. Cloud computing does the same for computing: instead of buying, running, and maintaining servers, you "plug in" to AWS or Azure and pay per second of computing time, per GB of storage, per million API calls. The complexity of data centres disappears.
Step by Step
Key Concepts
Virtual Machine (VM)
A software emulation of a physical computer, running on a hypervisor. Each VM has its own OS, CPU allocation, RAM, and storage, fully isolated from other VMs on the same host. AWS EC2 instances, Google Compute Engine VMs, and Azure Virtual Machines are all VMs. Slower to start than containers but provide stronger isolation.
Hypervisor
Software that creates and manages Virtual Machines, abstracting hardware resources and allocating them to VMs. Type 1 (bare metal): runs directly on hardware (VMware ESXi, Xen, KVM) — used in data centres. Type 2 (hosted): runs on a host OS (VirtualBox, VMware Workstation) — used for developer environments. AWS's Nitro hypervisor is a custom, stripped-down KVM variant optimised for cloud performance.
Availability Zone (AZ)
An isolated location within a cloud region — a separate data centre with independent power, cooling, and networking. AWS regions have 2-6 AZs. Deploying across multiple AZs protects against single data centre failures. Best practice: run critical applications in at least 2 AZs with load balancing and automatic failover.
Object Storage
Cloud storage for unstructured data (files, images, videos, backups) — not a filesystem. AWS S3, Google Cloud Storage, Azure Blob Storage. Data is stored as objects with unique keys in flat namespace buckets. Infinitely scalable (no provisioning), 11-nines durability (99.999999999% — one object lost per 100 billion stored per year), and priced per GB stored plus per-request fees.
CDN (Content Delivery Network)
A globally distributed network of edge servers that cache copies of content (images, videos, JavaScript files) close to users. When a user requests a file, it's served from the nearest edge node rather than the origin server — reducing latency from 200ms to <10ms for static content. AWS CloudFront, Cloudflare, Akamai. Essential for fast global web performance.
Shared Responsibility Model
Cloud providers secure "the cloud" (physical hardware, hypervisor, network infrastructure). Customers are responsible for securing what is "in the cloud" (data, access controls, application security, OS patching for IaaS). Misunderstanding this model is the leading cause of cloud security breaches — S3 buckets left publicly accessible, overly permissive IAM roles, unpatched OS.
FinOps
The practice of managing cloud costs — optimising spending without degrading performance. Cloud's pay-per-use model can lead to bill shock if resources are not rightsized (over-provisioned) or left running when idle. FinOps practices: reserved instances (commit to 1-3 year usage for 40-70% discount), spot instances (spare capacity at 60-90% discount), rightsizing, and eliminating unused resources.
Key Facts
- AWS (Amazon Web Services) generates more revenue than Amazon's entire retail business — $91 billion in 2023. Azure and Google Cloud generate $75 billion and $33 billion respectively. Together they account for ~65% of global cloud market share.
- Netflix runs entirely on AWS — 100% of its global streaming infrastructure. A single Netflix video stream involves dozens of AWS services: CDN, compute, databases, machine learning recommendations, and encoding pipelines.
- An estimated 90% of Fortune 500 companies use AWS. The US government's $10 billion JEDI cloud contract (won by Microsoft Azure) demonstrated that even the most sensitive government workloads are moving to cloud.
- The average enterprise uses 110 different cloud applications (SaaS tools) — up from 8 in 2015. "Multi-cloud" strategies using 2+ providers are now common to avoid vendor lock-in and optimise performance.
- Indian cloud market was worth ~$8 billion in 2023 growing at 30%+ annually. AWS, Azure, and Google have all established dedicated Indian regions (Mumbai, Hyderabad, Pune) serving compliance-sensitive Indian workloads.
- The carbon footprint of cloud data centres is significant but improving: hyperscalers like Google and Microsoft now run on 90%+ renewable energy. Cloud is typically 3-5x more energy-efficient than equivalent on-premise infrastructure due to higher utilisation rates.
Real-World Applications
Startup Infrastructure
Cloud enabled the lean startup era: a two-person startup can deploy a production-grade global application in hours with no capital expenditure. Stripe, Airbnb, Dropbox, and Slack all started entirely on AWS and scaled to billion-dollar companies without owning a single server.
Disaster Recovery
Traditional disaster recovery required maintaining a fully-equipped standby data centre — expensive to build and maintain. Cloud allows "warm standby" DR: replicate data to a cloud region and spin up infrastructure on demand only when disaster strikes. Recovery time drops from days to minutes; cost drops from millions to thousands per month.
Big Data and Machine Learning
Training large ML models requires thousands of GPUs — impractical to own. AWS SageMaker, Google Vertex AI, and Azure ML provide on-demand GPU clusters that can be rented by the hour, trained on, and released. What would have required a $50 million HPC facility a decade ago now costs a few thousand dollars on cloud.
Global Application Deployment
Cloud regions in 30+ countries allow applications to serve users from the nearest location — reducing latency and meeting data residency regulations. A fintech serving both Indian and European customers can keep Indian user data in AWS Mumbai and European data in AWS Frankfurt, complying with both RBI and GDPR data localisation requirements.
Remote Work Enablement
Google Workspace, Microsoft 365, Zoom, and Slack are all SaaS cloud applications that made the COVID-19 remote work transition possible. The shift from on-premise Exchange servers and VPN-accessed file shares to cloud-native collaboration tools was already underway but accelerated dramatically in 2020.
Frequently Asked Questions
Is the cloud secure?
Cloud providers invest billions in security — physical security, encryption, compliance certifications (SOC2, ISO27001, PCI-DSS, FedRAMP). Major cloud platforms have strong security track records for their infrastructure. The majority of cloud breaches are caused by customer misconfiguration — public S3 buckets, weak IAM policies, exposed API keys. The cloud shared responsibility model makes customers responsible for securing their data and access controls. Properly configured cloud environments are generally more secure than typical on-premise setups.
What is the difference between cloud and a regular web server?
A traditional web server is a fixed physical machine — limited CPU and RAM, requires manual setup, takes days to provision, and is wasted when idle. Cloud computing provides: instant provisioning (seconds via API), elastic scaling (automatically add/remove capacity), managed services (databases, caches, queues you don't maintain), global availability (deploy to 30 regions instantly), and pay-per-use billing. The same application running on both might be indistinguishable to users, but cloud has enormous operational advantages.
What happens to my data if AWS goes down?
AWS outages do occur — typically affecting a single Availability Zone or service within a region. Well-architected applications are designed for this: data is replicated across multiple AZs, traffic is load-balanced, databases have automatic failover, and backups exist in a separate region. During the major AWS us-east-1 outage in 2021, many websites relying solely on that region went down — demonstrating why multi-AZ and multi-region architectures matter for critical applications.
How does cloud pricing work?
Cloud pricing has dozens of dimensions: compute (per vCPU-hour), memory (per GB-hour), storage (per GB-month), network (per GB transferred out), API calls (per million requests), managed services (per unit of usage). AWS alone has 200+ services each with their own pricing model. The pay-per-use model is powerful but can lead to unexpected bills from forgetting to stop resources or from traffic spikes. Tools like AWS Cost Explorer, reserved instances, and automated budget alerts help manage cloud spend.