What is Cloud Computing?

Renting computing resources over the internet, on demand, and paying only for what you use.

What is it?

Cloud computing means using someone else's data centers - servers, storage, networking, databases - through the internet, instead of buying and running your own hardware. You ask for what you need, you get it in minutes, and you are billed for what you actually consumed.

Almost everything on the web follows the client-server model: a client (your browser or phone app) sends a request, and a server (a computer somewhere) sends back a response. Cloud providers like AWS run enormous fleets of those servers and let you rent slices of them.

Three ideas define the cloud:

  • On-demand self-service: you provision resources yourself through a console or API, with no purchase order and no waiting for a delivery truck.
  • Pay-as-you-go: no large upfront purchase; the meter runs only while you use something.
  • Elasticity: you can grow or shrink capacity as demand changes.

Deployment models describe where the infrastructure lives:

  • Cloud: everything runs on a cloud provider. You can build there from scratch or migrate existing apps.
  • On-premises (also called a private cloud when it offers self-service on your own hardware): everything runs in your own data center.
  • Hybrid: a mix. Some workloads stay in your data center (for latency, regulation, or because they are hard to move) while others run in the cloud, connected by a network link.

AWS commonly lists six advantages of cloud computing:

  • Trade upfront (capital) expense for variable expense.
  • Benefit from massive economies of scale.
  • Stop guessing capacity.
  • Increase speed and agility.
  • Stop spending money running and maintaining data centers.
  • Go global in minutes.

Cost and licensing ideas that show up on the exam

  • Fixed vs variable costs: fixed costs (buying servers, leasing a building) are paid whether or not you use them; variable costs (cloud usage) rise and fall with consumption. The cloud shifts spending from fixed to variable.
  • License models: License Included means the software license is bundled into the hourly price (for example Windows Server or SQL Server on EC2/RDS). Bring Your Own License (BYOL) means you reuse a license you already own, which can lower cost if your contract allows it.
  • Rightsizing: matching instance types and sizes to what a workload really needs, so you do not pay for idle CPU or memory. Measure first, then shrink or change type.
  • Managed services: AWS runs the underlying servers, patching, backups and scaling for you (RDS, DynamoDB, Lambda), so your team spends less effort on operations and more on the product. The trade-off is less low-level control.

Automation is a core cloud benefit: infrastructure as code tools such as AWS CloudFormation let you describe resources in a template and create, update and delete whole environments the same way every time. Repeatable stacks reduce manual mistakes, speed up recovery, and make it cheap to build a copy for testing.

Explain like I'm 10

Owning your own servers is like buying a truck to move apartments once a year: you pay for it, insure it, park it, and it sits idle most of the time. The cloud is like a van-rental app - you tap, a van shows up, and you pay for the hours you drove. Need a bigger vehicle for a week? Tap again.

Examples

On-premises vs cloud: the same need, two ways

Need: a web server for a product launch next month

On-premises:
  1. Forecast peak traffic (guess)
  2. Order hardware, wait weeks for delivery
  3. Rack it, cable it, install the OS
  4. Pay for it whether or not anyone visits
  5. After launch it sits mostly idle

Cloud:
  1. Launch a server (an EC2 instance) in minutes
  2. Add more during the launch spike
  3. Remove them afterwards
  4. Pay only for the hours they ran

The work is the same; the difference is when you pay and how fast you can change your mind.

Your first cloud request, from the terminal

# Ask AWS which account and identity you are using
aws sts get-caller-identity

# List the Regions your account can use
aws ec2 describe-regions --query "Regions[].RegionName" --output text

Every cloud action is just a request to an API. This is the smallest useful one: 'who am I?'

How it works

A cloud provider builds huge data centers and uses virtualization to slice physical machines into many isolated virtual ones, so thousands of customers can safely share the same hardware (multi-tenancy). Software on top exposes every resource as an API call: create a server, attach storage, open a network port.

Because the provider buys hardware at enormous volume, its cost per unit falls - that is the economy of scale behind lower prices. And because capacity is pooled, you can scale up for a spike and release it when finished, so you stop guessing capacity months ahead.

Hybrid setups usually connect the two worlds with a VPN or a dedicated network line, so your on-premises systems and cloud systems can talk as though they were on one network.

   CLIENT                          CLOUD PROVIDER
 +---------+   request    +-----------------------------+
 | browser |------------->|  Region                     |
 | / app   |<-------------|   +-------+  +-----------+  |
 +---------+   response   |   |server |  | database  |  |
                          |   +-------+  +-----------+  |
   You rent these ------->|   (virtualized, pooled)     |
   and pay by usage       +-----------------------------+

Why does it exist?

Running your own data centers means big upfront spending, long lead times, and capacity that is either too small at peak or too large the rest of the year. Small teams could not afford infrastructure that large companies took for granted.

Cloud computing turns infrastructure into a utility, like electricity: you do not build a power plant to run a lamp. That lets a two-person team experiment cheaply and a large company expand to a new continent without building anything.

When to use it

Choose the cloud when demand is uncertain or spiky, when you want to launch quickly, when you need to reach users in many countries, or when you would rather spend engineering time on your product than on racking servers.

Choose hybrid when regulation, latency to local equipment, or a large existing investment keeps some workloads on-premises.

When not to use it

A steady, predictable, very large workload on fully depreciated hardware may be cheaper to keep on-premises. Strict data-residency rules or equipment that must sit next to a factory floor may also force on-premises or hybrid. The cloud is a tool, not a moral position - compare real costs.

Common mistakes

  • Thinking 'cloud' means 'someone else handles all security' - you still own your data, identities, and configuration (see the shared responsibility model).

  • Assuming the cloud is always cheaper; unused resources left running still cost money.

  • Confusing a private cloud with plain on-premises servers; a private cloud adds self-service and elasticity.

  • Forgetting that on-demand cuts both ways: resources you forget to turn off keep billing.

  • Treating hybrid as a temporary failure rather than a legitimate long-term design.

  • Treating BYOL as always cheaper - it only helps if your existing license terms allow use on shared cloud hardware.

  • Rightsizing once at launch and never again; usage changes, so review it regularly.

Practice exercises

  1. Easy:

    List the six advantages of cloud computing from memory, and give a one-line example of each.

  2. Easy:

    A news site gets 50x traffic during elections and little otherwise. Explain which cloud benefits apply and why on-premises would struggle.

  3. Medium:

    A bank must keep customer records in its own data center for regulatory reasons but wants to run analytics in the cloud. Sketch a hybrid design and name what connects the two sides.

  4. Medium:

    Run aws sts get-caller-identity (or read its output in the docs) and explain each field it returns.

  5. Hard:

    Build a rough 3-year cost comparison for a workload that needs 20 servers year-round vs one that needs 20 servers for two months a year. Where does the cloud win and where might it not?

Interview questions

What is cloud computing?

On-demand delivery of IT resources over the internet with pay-as-you-go pricing, instead of owning and running physical infrastructure yourself.

Name the three cloud deployment models.

Cloud (all in the provider), on-premises (all in your own data center, sometimes called private cloud), and hybrid (a connected mix of both).

What does 'trade upfront expense for variable expense' mean?

Instead of buying hardware (capital expense) before you know demand, you pay per use as an operating expense, so cost tracks actual consumption.

Exam-style: A company wants to stop guessing how much server capacity it will need. Which cloud benefit addresses this? (A) Economies of scale (B) Stop guessing capacity (C) Go global in minutes (D) Stop running data centers

B. You can scale up or down as demand changes instead of forecasting months ahead.

Exam-style: Which model keeps some workloads in a company data center and others in AWS, connected together?

Hybrid deployment.

Exam-style: Which statement describes a variable cost? (A) Buying a server (B) Paying per hour for compute (C) Leasing a data center (D) Buying cooling equipment

B. Variable costs scale with usage; the other three are fixed.

What is the benefit of a managed service?

AWS handles infrastructure maintenance such as patching, backups and scaling, reducing the customer's operational effort.

What does CloudFormation automate?

Creating and managing a set of AWS resources from a template, so environments are repeatable.