1.2 Where the cloud came from

Describes the cloud landscape as of August 2026

What this is and why it exists

The cloud looks new and is not. Knowing where it came from tells you why it is shaped the way it is — and why the same idea failed a decade before it worked. This lesson traces the line from time-shared mainframes to a rented GPU.

The vocabulary

Time-sharing — many users at terminals sharing one expensive mainframe, each behaving as if it were theirs.
Client-server — the 1980s-90s reversal: cheap machines on desks, work moved out of the centre.
Grid computing — many machines pooled to work on one large job.
Utility computing — the idea that computing should be metered and sold like power or water.
Autonomic computing — systems that manage, heal and tune themselves with minimal human care.
Service-oriented architecture (SOA) — software built as services that call each other over a network.
Virtualization — one physical machine safely divided into many isolated virtual ones.

The mental model

A pendulum. Computing began centralised (the mainframe, because hardware was scarce and dear), swung out to the desktop (client-server, because hardware got cheap), and swung back to the centre (cloud, because scale got cheap and networks got fast). Each swing kept the previous era's lessons: the cloud is time-sharing again, at planetary scale, with the four roots supplying its organs — grid gave pooling, utility gave metering, autonomic gave self-management, SOA gave the service interface.

Around 2006 three things landed together and made it real: bandwidth became cheap enough to move work over the network, commodity hardware became cheap enough to fill warehouses, and virtualization finally worked well enough to share one machine safely among strangers. That third one is the keystone — without isolation, pooling is a security incident.

The application service provider model of 1999 was the same pitch — rent software over the network — and it failed because none of the three had landed: bandwidth was dear, each customer needed dedicated machines, and there was no self-service. Same idea, missing ingredients.

What you should now be able to explain or do

Draw the pendulum in words: centralised, decentralised, centralised again, and what pushed each swing. Name the four roots and the organ each gave. Explain in two sentences why 1999's version died and 2006's lived.

Check yourself

What did virtualization contribute that made pooling possible?

Safe isolation — one physical machine divided into virtual ones that cannot see each other, so strangers can share hardware without sharing data.

Name the four roots of cloud computing and one thing each contributed.

Grid computing — pooling many machines; utility computing — metered billing; autonomic computing — self-managing systems; SOA — services callable over a network.

Why did the application service provider model fail in 1999 when the same idea worked in 2006?

The ingredients were missing: bandwidth was expensive, hardware was not commodity, virtualization was immature — so no pooling, no elasticity, no self-service economics.

In what sense is the cloud "time-sharing again"?

Many users sharing one centralised pool of expensive infrastructure, each experiencing it as their own — the mainframe pattern at planetary scale.

Go deeper

We haven't checked most of these for screen reader use yet.

Back to Where the cloud came from: work through the checklist