The case

Bought for one day. Idle on all the others.

Isaac sells specialty coffee online. To survive Black Friday, he bought servers for the biggest peak of the year. On the other days, almost all of it sits idle, but the bill keeps coming.

capacity bought for the peakBlack Friday ↓idle (and paid for) the rest of the yearjanfebmaraprmayjunjulaugsepoctnovdec

The problem

Three situations, the same pain

Online store

Buys for Black Friday and sits idle the rest of the year.

When the app takes off

From a thousand to a hundred thousand users in one week. Buying a server takes weeks.

International expansion

A data center in another country takes months and costs a fortune.

Before the cloud, every company bought and maintained its own servers. A server has limited capacity and is a single point of failure, and a data center needs power, cooling, networking and security 24 hours a day.

The solution

Rent instead of buy

Nothing in the cloud happens by enchantment: the servers belong to a provider, and you pay for the time you use them. Three characteristics define the model:

On-demand

Resources right away, with no purchase order.

Elasticity

Capacity goes up and down with demand.

Pay-as-you-go

You pay only for what you used.

Own data centerCloud
Green: real demand · Dashed: available capacity · Red: what is left over

Back to Isaac: with the cloud, he spins up 20 servers on the Friday of Black Friday and goes back to 2 on Monday.

Benefits

What you get

  1. 01

    CapEx → OpEx

    A fixed investment becomes spending as you use.
  2. 02

    Economies of scale

    The provider buys wholesale and passes the discount on.
  3. 03

    Elasticity

    It scales on its own, with no idle resources.
  4. 04

    Agility

    A new server in minutes, not months.
  5. 05

    Global reach

    Regions all over the world, less latency.
  6. 06

    No heavy lifting

    The infrastructure stays with the provider.

Key terms

Six words to remember

These terms show up all the time in any cloud certification, whether AWS, Azure, Google Cloud or Oracle Cloud.

Tap a card to see the definition.

ArchitectureFinancial

Don't mix them up

Elasticity × Agility

Both sound like "responding fast". The difference lies in the question each one answers.

Elasticity

“Do I have enough resources right now?”

Example
A streaming channel adds servers on the night of the national team match and reduces them the next day.
Focus
Capacity following the load.

Agility

“How fast do I launch something new?”

Example
A team creates a test environment in 5 minutes to validate a feature.
Focus
Speed of innovation.

CapEx × OpEx

CapEx: everything on day 1OpEx: follows usagetime →accumulated spending
Buying a server (CapEx) × renting a VM by the hour (OpEx)

It is OpEx that lets a team start small and grow without spending all its capital right at the start.

Behind the benefits

Three ideas that explain the rest

Economies of scale

It is buying rice wholesale instead of at the corner shop, only on a planetary scale. The provider pays less per unit and passes part of the discount on. That is why cloud prices tend to fall over time.

Undifferentiated heavy lifting

Heavy work that has to be done, but that doesn't set you apart from your competitors. Think of a restaurant:

Wins customers with

  • The menu and the recipes
  • Service and the atmosphere

Doesn't win customers with

  • Gas and the kitchen's electrical wiring
  • Maintenance of the stove and the cold room

Peak load provisioning

It is Isaac's chart from the beginning: buying for the expected peak and leaving the capacity idle the rest of the time. Elasticity solves this, scaling up at the peak and going back to the minimum (or to zero) afterwards.

Quick test

Three questions, exam style

01

A store's traffic is 10× higher in December. Which characteristic avoids paying for idle servers the rest of the year?

02

A team creates a test environment in 5 minutes to validate an idea. That is…

03

The company signs a 3-year usage commitment in exchange for a discount. This spending looks more like…


These concepts hold for any cloud certification: AWS, Azure, Google Cloud or Oracle Cloud. This post is the common starting point: each certification's posts start here and go on to their provider's services and terms.