What is Cloud Computing?
Utility computing for the modern era
Cloud computing is the on-demand delivery of compute, storage, networking and software over the internet, billed like a utility instead of bought like hardware.
What you will learn
- Define cloud computing in terms of the NIST essential characteristics
- Explain the on-demand, pay-as-you-go model versus buying data centers
- Name the major providers and the kind of platform each offers
Instead of buying servers you may never fully use, you rent exactly the compute, storage and networking you need — by the hour or second — from a provider that runs vast, highly automated data centers.
What is this?
Cloud computing is the delivery of computing resources — servers, storage, databases, networking, software and analytics — over the internet on demand, with pay-as-you-go pricing. The NIST definition rests on five characteristics: on-demand self-service, broad network access, resource pooling, rapid elasticity and measured service. AWS, Azure and Google Cloud dominate the space, but the model applies equally to smaller and specialized providers.
Why do you need it?
- Buying a data center is a years-long commitment made on guesses; the cloud lets you pay for what you actually use.
- Elasticity means a viral launch goes from 10 servers to 10,000 and back without a procurement ticket.
- Providers operate at a scale that makes isolation, security and reliability someone else's full-time problem.
- The operational lessons of the cloud — automation, ephemerality, tagged resources — apply to on-prem too.
Mental model
Buying a car vs taking a ride-hailing app
Owning a data center vs cloud computing
The car (data center) means capital, maintenance, insurance and depreciation — and it sits idle most of the day. The ride-hailing app (cloud) is there when you need it, you pay per trip, and the fleet belongs to someone else.
How it works
- Providers build hyperseale data centers with pooled CPU, memory, storage and networking resources.
- A control plane (web console, CLI, API) gives customers self-service access to allocate resources in minutes.
- Virtualization and orchestration slice physical hardware into isolated, resizable units per customer.
- Metering tracks usage per resource per second or per GB, enabling utility-style billing with no upfront cost.
- Consumers interact through the same abstraction everywhere: the API is the product.
The modern cloud computing model.
Examples
On-demand provisioning, without touching a physical server
# Check whether you could launch a VM (dry-run = no cost)aws ec2 run-instances --image-id ami-0abcdef1234567890 \ --instance-type t3.micro --dry-run# An error is EXPECTED here — it proves the API call is authorized.The dry-run flag submits the request but stops before creating anything billable. The 'DryRunOperation' error is the API saying 'yes, you have permission and this would succeed'. It is the single safest way to prove your credentials and permissions work.
In production
- A startup launches its first product with a few VMs and a managed database, then scales to autoscaling fleets as traffic grows — all without buying hardware.
- A bioinformatics lab bursts from 50 CPUs to 20,000 spot CPUs for a weekend genome-sequencing run, then scales back to zero.
- A regional bank keeps its regulated on-prem core but moves analytics, DR and burst workloads to the cloud, treating the cloud as utility capacity.
Best practices
- Automate everything from day one — the console is for exploration, the API is for production.
- Tag every resource so you can attribute cost to a team, environment or project.
- Scope credentials to least privilege even before you have users to protect.
- Decide a cloud-finance model (per-team budgets, chargeback) before the bill surprises you.
Common mistakes
- Recreating the physical data center mindset (long-lived 'pets') in a world built for automation.
- Treating the cloud as 'somebody else's computer' and skipping backups and IAM hardening.
- Assuming on-demand pricing is the cheapest — sustained use needs reservations and rightsizing.
- Building one giant monolith machine instead of accepting the cloud's elasticity model.
Common issues
CLI says 'Unable to locate credentials'No default credential chain is configured on the machine
Run aws configure (or set the provider's env vars/ADC file) before making API calls.
Instance starts but has no networkThe VM was launched outside any VPC/subnet with no internet route
Launch into a VPC with an internet gateway and correct route table; check security groups.
Surprise end-of-month billResources (instances, volumes, IPs) left running with no tags or lifecycle
Add budgets + alerts and a scheduled stop/terminate policy for non-production environments.
Debugging tips
- Begin any cloud diagnosis with the IAM check: 'are my credentials valid and scoped to this action?'
- Use read-only calls (describe/get/list, --dry-run) before any mutation to confirm assumptions.
- Reproduce the problem in a second region or account to separate provider incidents from your config.
- Read the API error code, not just the message — 'AccessDenied' and 'Timeout' need different hunts.
Interview checkpoint
Hands-on lab
- Model the cost difference between on-demand cloud usage and buying hardware
- See how utilization drives the right economic decision
- Practice reasoning about cloud vs capex trade-offs
Simulate Pay-as-you-go vs Buying Servers
25 min · beginner
Quick quiz
Which NIST characteristic means a customer can provision resources without human intervention from the provider?
The cloud guarantees your total cost will always be lower than on-premises hardware.