Diagramium
☁️ Narrated diagram

Cloud computing: IaaS, PaaS and SaaS

Start in your own server room, then climb the ladder — infrastructure, platform, software and serverless — with a real product and a clear owner at every rung.

16 steps3 min readNarrated with a studio voice
Cloud computing: IaaS, PaaS and SaaSalways sits onkeep all nine yourselfwhich meanshand over the hardwaresuch ashand over the platformsuch ashand over the whole appsuch ashand over the idle timesuch asthe line moves withbilled byremembered asinformsinformsexplainsA key idea or topic — the main building block of the mapCloud computingA narrower idea branching off a conceptThe nine layer stackA narrower idea branching off a conceptOn premisesA narrower idea branching off a conceptYou manage all nineA narrower idea branching off a conceptInfrastructure as a serviceA narrower idea branching off a conceptElastic Compute CloudA narrower idea branching off a conceptPlatform as a serviceA narrower idea branching off a conceptHeroku and App EngineA narrower idea branching off a conceptSoftware as a serviceA narrower idea branching off a conceptGmail and SalesforceA narrower idea branching off a conceptServerless functionsA narrower idea branching off a conceptLambda and Cloud RunA narrower idea branching off a conceptShared responsibilityA narrower idea branching off a conceptPay for what you useA narrower idea branching off a conceptThe pizza testA concrete example of a concept — drawn as an ovalChoosing your rung
Cloud computing: IaaS, PaaS and SaaS — the complete diagram. Press Present to watch it build itself.
Step 1 of 16

Cloud computing

Cloud computing is renting somebody else's computers, but the interesting question is how much of the work you rent along with them. The answer splits into a few named rungs, and those names are on almost every technology job advert written today.

Step 2 of 16

The nine layer stack

Running any application means somebody handles nine things. The building, the power, the servers, the storage, the network, the virtual machines, the operating system, the runtime and the application itself. Your data sits on top of all nine, and it is always yours.

Step 3 of 16

On premises

Before the cloud, all nine of them were yours. You bought a rack, air conditioned a room, and ordered machines twelve weeks ahead based on a guess about next year's traffic. Guess low and you crash; guess high and you have bought idle metal.

Step 4 of 16

You manage all nine

The hidden cost was never the hardware, it was the people. Somebody had to patch operating systems at two in the morning and swap failed disks, and a typical company server sat idle at under fifteen percent of its capacity.

Step 5 of 16

Infrastructure as a service

Infrastructure as a service hands you the bottom layers. You rent virtual machines, disks and networking by the hour, and you still install and patch the operating system yourself. Maximum control, and maximum homework to go with it.

Step 6 of 16

Elastic Compute Cloud

Amazon's Elastic Compute Cloud is the archetype, alongside Azure Virtual Machines and Google Compute Engine. A server that once took three months to purchase now takes about forty seconds to start, and you can delete it just as quickly.

Step 7 of 16

Platform as a service

Platform as a service takes the operating system away as well. You hand over code, and the provider handles the machines, the patching and the scaling. The trade is convenience for control, and you build the way the platform prefers.

Step 8 of 16

Heroku and App Engine

Heroku made this famous with a single command that pushed a web application live. Google App Engine, Azure App Service and Vercel do the same job today, and most developers using them never learn which machine their code is running on.

Step 9 of 16

Software as a service

Software as a service is the top of the ladder. There is nothing to install and nothing to operate. You open a browser tab and use finished software, usually paying a fee per person per month and nothing else.

Step 10 of 16

Gmail and Salesforce

Gmail, Salesforce, Slack and Dropbox all live here. You manage exactly one layer, your own data, and that happens to be the layer worth thinking hardest about, because getting your data back out again is often the difficult part.

Step 11 of 16

Serverless functions

Serverless goes one rung further. You upload a single function and it runs only when something calls it, scaling from zero to thousands of copies and back down again within seconds. Sitting idle costs you nothing at all.

Step 12 of 16

Lambda and Cloud Run

Amazon's Lambda bills in slices of a single millisecond, so a function called a hundred times a day can cost less over a year than one cup of coffee. The catch is the cold start, the small pause while a sleeping function wakes up.

Step 13 of 16

Shared responsibility

Every provider publishes a shared responsibility model, and it repays reading carefully. They secure the cloud itself; you secure whatever you put inside it. Many of the loudest cloud leaks were storage buckets left open to the public internet, not broken data centres.

Step 14 of 16

Pay for what you use

The deeper change is the billing clock. A large capital purchase became a running operating cost, so three people in a room can start a hundred servers for one afternoon experiment and hand them all back before dinner.

Step 15 of 16

The pizza test

The classic way to remember the ladder is pizza. Cooking at home is on premises, a take and bake tray is infrastructure, delivery to your door is platform, and sitting down in the restaurant is software as a service.

Step 16 of 16

Choosing your rung

Higher up means less work and less control, and the right answer is almost never one rung for everything. Sensible teams buy finished software for email, use a platform for the web app, and rent raw machines only where something genuinely unusual lives.

Watch it explain itself

Every step above is narrated aloud. Play it, or open it in the editor and make it yours — no account needed.