How cloud computing works.

A cloud service runs on real computers in buildings called data centres, which the provider owns and operates. You reach them over a network. Software divides each machine between customers, and a meter records what each customer uses.

Data centres and regions

Providers group their data centres by location and let you choose where your systems run. AWS, for example, calls a physical location a Region. Each Region contains several Availability Zones, and each zone is one or more separate data centres with its own power and network connections. Microsoft describes an Azure region in similar terms: one or more data centres joined by a fast network.

The choice of region decides two things. The first is how well a system survives a fault: AWS says spreading a system across zones makes it more fault tolerant than running it in a single data centre. The second is which country holds your data. The Australian Signals Directorate (ASD) advises using only cloud services located in countries that suit the sensitivity of the data.

One machine, many customers

A provider does not give each customer a physical server. It uses virtualisation: software that lets a single server run many separate operating systems at once and keeps each one apart from the others. NIST describes this as the technique that makes more efficient use of hardware and allows many customers to share it.

That sharing is the pooling in NIST’s definition of cloud computing. The provider’s equipment is assigned and reassigned as demand moves, which is why you can ask for a new server and have it quickly, with nobody at the provider involved.

Paying for what you use

Cloud systems measure use automatically. Depending on the service, the meter counts storage, processing, network traffic or active user accounts, and NIST notes that charging is typically based on it. Capacity can also grow and shrink with demand.

The practical effect is a bill that moves. Review it each month, and ask for an alert when spending passes an amount you set.

Who secures what

ASD calls the split of security duties the shared responsibility model. The provider is responsible for services that are securely designed, set up and operated, including the infrastructure underneath. You are responsible for your data and who can reach it, the devices you use, the access you give people, any software you bring, and acting on security alerts.

Where the line falls depends on the service. If you rent a virtual server, AWS states that looking after its operating system is the customer’s job, updates included. With a finished application, the provider does far more.

What to ask a provider

• Which region will hold our data, and can we fix it to Australia?
• Where is your shared responsibility document for this service? ASD suggests looking elsewhere if a provider has none.
• Do you back up our data, or do we?
• Will you alert us to unfamiliar sign-ins, and does that cost extra?

If you are weighing a move, read Is it time to move your business servers to the cloud?. Our cloud services page covers the setup and running.