PE2-6.1 Introduction to Cloud Computing

What this is and why it exists

The word cloud is used loosely enough to mean almost nothing. This topic gives it a definition sharp enough to be tested against a vendor's description.

There is always someone else's computer underneath. What makes it a cloud is a specific set of properties, and something lacking them is renting a server with better marketing.

The vocabulary

• On-demand self-service — obtaining resources without asking a person.
• Broad network access — reaching them over the network from ordinary devices.
• Resource pooling — many customers sharing one set of physical machines.
• Rapid elasticity — capacity growing and shrinking quickly with demand.
• Measured service — usage metered, so you pay for what you use.
• Reliability — how seldom a service fails.
• Availability — the fraction of time it is working.
• Serviceability — how quickly it can be repaired when it is not.
• Vendor lock-in — depending on one provider so heavily that leaving is expensive.
• Data residency — the requirement that data stays within a particular country.

The mental model

The five characteristics are the definition, and they work best as a checklist.

Can you obtain a resource yourself, immediately, without emailing anyone? Can you reach it over an ordinary network? Is the underlying hardware shared with other customers? Can capacity grow and shrink quickly with demand? Is usage measured so you pay only for what you used?

Answer no to several of those and it is hosting. That checklist is more useful than any definition, because it can be applied to what a vendor actually offers.

The origins are worth a moment, because most of these ideas predate the name by decades. Timesharing systems shared one expensive machine among many users. Grid computing pooled machines across institutions. Utility billing charged for consumption rather than ownership. What changed was the scale, the automation and the price, not the concept.

Two benefits are genuinely new rather than repackaged. Elasticity means capacity can follow demand within minutes, so you do not buy for the peak and pay for it all year. Paying for what you use turns a capital purchase into an operating cost, which changes who can start something.

The risks belong beside the benefits from the beginning, not in a final paragraph.

Vendor lock-in is the hardest to reverse. Using a provider's own services rather than portable ones makes an application faster to build and much harder to move. That is a decision made early and paid for late.

Data residency is a legal constraint rather than a technical one. Some data must remain within a country's borders, and that limits where it may be stored regardless of what is convenient.

Loss of direct control is the one people feel most. When the service is down you cannot walk to the machine. You wait for someone else's engineers, on their schedule, and your customers wait with you.

Reliability, availability and serviceability give a vocabulary for judging a service, and they are the same three used in ordinary systems engineering. Reliability is how seldom it breaks. Availability is what fraction of the time it works. Serviceability is how fast it is restored. A service can be reliable and unavailable, if the rare failure takes days to fix.

The comparison with buying ordinary hosting is the practical frame. A hosting provider gives you a machine after a conversation and an invoice, and you keep it whether you need it or not. A cloud provider gives you the machine in seconds without a conversation, and takes it back when you stop paying for it. Self-service and elasticity are the difference.

What you should now be able to explain or do

State the five characteristics and use them as a checklist against a vendor's description. Say which ideas predate the term and what actually changed. Name the two genuinely new benefits. State the three risks that are hard to reverse and say why each is hard. Distinguish reliability, availability and serviceability.

Check yourself