What Cloud Computing Is (and Isn't)

A question at the packing bench

Pennyworth Crafts is a small online shop that sells yarn, paper, paints and craft kits. Its owner, Beth Carrington, has just hired Sophie Clarke as a junior cloud administrator, and on Sophie's first morning Beth asks the question many business owners ask sooner or later.

“Everyone tells me to ‘move to the cloud’,” she says. “My accountant says it will save money, my nephew says it is just someone else's computer, and the man at the trade fair said nothing ever goes down in the cloud. What actually is it?”

This course is the answer to Beth's question, told through Sophie's first months in the job. Along the way you will move Pennyworth Crafts' shop, piece by piece, from the single computer in its back room into a cloud, and learn why each piece is done the way it is. You do not need any programming or earlier cloud knowledge. Some familiarity with using a computer and the web is enough.

Cloud computing in one sentence

Cloud computing means renting computing resources, such as machines, storage, networks and ready-made services, from a provider over the internet, setting them up yourself in minutes, and paying only for what you use.

Every part of that sentence matters:

• Renting, not buying. The provider owns the buildings, the computers, the cables and the electricity bill. You pay to use a share of them.
• Computing resources: not just storage for files. You can rent whole computers, private networks, databases, and services that do a particular job for you.
• Over the internet: you never see or touch the equipment. You control everything remotely, from a website, from typed commands or from files that describe what you want.
• Setting them up yourself in minutes: no order form, no delivery, no waiting for an engineer.
• Paying only for what you use, measured by the hour (or even the second), by the gigabyte or by the request, with the charge stopping once you remove what you rented.

A useful comparison is electricity. A hundred years ago, a factory that wanted electric power often built its own small power station. Today almost every business plugs into the mains, uses as much as it needs, and pays by the unit. Nobody thinks of that as “losing control of their power station”; they think of it as not having to run one. Cloud computing does the same for computers: instead of buying, housing and maintaining machines, you plug into a provider and pay by the unit. The comparison is not perfect, as you will see, but it is a good place to start.

Five features of a cloud service

A widely used definition of cloud computing lists five features. A service that has all five is a cloud service; one that lacks several of them (a single rented machine you must order by email and pay for by the year, say) is ordinary hosting. Here they are, with what each would mean for Pennyworth Crafts.

Compare that with the office server. Getting a second one means ordering it, waiting for delivery, finding a shelf and a power socket, and installing it: days or weeks, not minutes. It is paid for whether it is busy or idle. When the sale ends, it cannot be handed back. None of the five features apply.

Where is the cloud, really?

The name makes it sound vague and far away, but the cloud is very physical. A cloud provider owns data centres: large, secure buildings full of racks of computers, disks and network equipment, with their own power supplies, backup generators and cooling. The word “cloud” comes from old network diagrams, where engineers drew the internet, and anything else they did not want to draw in detail, as a cloud shape.

Providers group their data centres by location. A region is one geographical area, often around a large city, that contains several data centres; big providers have dozens of regions around the world. Within a region, the data centres are grouped into zones: separate sites far enough apart that a fire, flood or power cut at one should not affect the others, but close enough to be joined by very fast private connections. The practice cloud in this course has two regions, North and South, each with three zones. Lesson 2.4 explains how to use them, and module 5 shows how zones keep a shop running when a whole building fails.

Inside a data centre, each physical computer is usually split into several virtual machines: separate, complete computers made by software, each with its own share of the processors, memory and disk, and its own operating system. When you “create a server” in a cloud, you almost always get a virtual machine. You choose its size, and the provider decides which physical machine it runs on. Lesson 2.1 looks at this more closely.

What the cloud is not

Beth heard three confident claims. Each contains some truth, and each can lead to expensive mistakes if taken at face value.

“It's just someone else's computer”

True in one sense: your servers do run on the provider's hardware. But it misses what you are really renting. A single computer in someone else's building would have none of the five features: you could not create ten more of them in a minute, spread them across three buildings, or pay by the hour. What makes the cloud useful is the huge pool of equipment, the automation that hands it out, and the ready-made services built on top.

“It will save money”

Sometimes. The cloud is excellent value for work that varies (a shop with busy sale weeks and quiet months), for new projects that may not last, and for small teams who would otherwise spend their time looking after hardware. It can be more expensive than owning equipment for work that is steady, large and predictable, and it is very easy to waste money on resources that are forgotten or bigger than they need to be. Module 8 is about keeping the bill under control.

“Nothing ever goes down in the cloud”

Not true. Providers' equipment fails every day; they are simply very good at limiting the damage. Individual servers stop, disks break, occasionally a whole zone has trouble, and once in a while a problem spreads across a whole region. A shop that runs on one cloud server in one zone can go down just as easily as the office server did. Staying up is something you design for, using several servers in several zones, as you will do in module 5.

Public, private and hybrid

When people say “the cloud” they usually mean a public cloud: a provider that rents its resources to anyone, over the internet. That is what this course is about. A private cloud uses the same ideas (self-service, pooling, automation) on equipment used by only one organisation, often in its own data centre; it suits very large organisations with special rules about where their data may be. A hybrid setup mixes the two, or mixes a public cloud with ordinary equipment the organisation already owns. Pennyworth Crafts will spend a short time as a hybrid on moving day in module 10, when some of the shop is in the cloud and some is still on the office server.

Buying or renting

The biggest change the cloud brings is to how computing is paid for. Buying equipment is a large payment up front for something you then own and must look after for years; accountants call this capital expenditure. Renting is a smaller, regular payment that follows how much you use; this is operating expenditure. Neither is always better.

For Pennyworth Crafts the balance is clear. The shop is quiet most of the year and very busy for a few weeks. Its one machine is already too small for the busy weeks, and a power cut in the back room shuts the business. Renting lets it pay for a small shop most of the time and a big one only when it needs it, spread across buildings that have their own backup power. The rest of this course shows how.

A glimpse of what is coming

In the next lesson you will meet the practice cloud, a pretend cloud provider that runs inside your web browser, where you will do all the hands-on work in this course. Here is a preview. The first practice cloud below is the shop as it is today, on the office server, facing a busy sale day of 90 requests a second. Press the Try buttons in order and read the status.

The office server is overloaded, and a third of the visitors' requests fail. The second practice cloud holds a copy of the shop that Ian has already built in the cloud, on two servers behind a load balancer (lesson 5.1), with a managed database (lesson 6.1). The same 90 requests a second are sent to it as a test.

Nothing fails, and a page takes less than a fifth of a second on average. It is not free, though: the bill shows every part of the cloud copy with its cost per hour, and the servers are already working hard. Doubling the traffic would need more servers, which is exactly what you will learn to add, automatically, in lesson 5.2. Do not worry about the details yet: by the end of the course you will be able to build all of this yourself, explain every line of that bill, and make it smaller.

Quick check

Choose an answer to see whether you are right and why.

Found a mistake on this page, or something unclear? Report a problem and mention “Cloud Computing Basics Lesson 1.1”.