PE2-6.2 Virtualization & Service Level Agreements
Standard cloud computing concepts — written September 2026
What this is and why it exists
One technology made the economics work. Without it, a provider would need one physical machine per customer and the price would be no better than owning it yourself.
Virtualization lets one machine be safely divided into many. That single capability is what everything else rests on.
The vocabulary
- Virtualization — presenting one physical machine as several independent ones.
- Hypervisor — the software layer creating and separating those machines.
- Multi-tenancy — several customers sharing one physical machine.
- Utilisation — the fraction of a machine's capacity actually being used.
- Shared responsibility — the division of security duties between provider and customer.
- Service level agreement — the contract stating what the provider promises.
- Availability target — the fraction of time the service promises to be working.
- Service credit — the usual remedy when that promise is missed.
The mental model
Start with the economics, because they explain the technology.
A physical server bought for a workload spends most of its life idle. Peak demand arrives for a few hours a day, and the machine is sized for the peak. Average utilisation of a few per cent is ordinary.
Virtualization changes that. Put several workloads on one machine, each in its own virtual machine, and the peaks of different customers fall at different times. Utilisation rises sharply and the cost per customer falls. The provider's whole business is that arithmetic.
The hypervisor is what makes it safe. It presents each virtual machine with what looks like its own hardware, and it keeps them from reaching each other's memory. That separation is the assumption the entire model rests on, and it is enforced by software, which is worth remembering when the security topic arrives.
The lineage is worth knowing. Distributed systems supplied the ideas about coordination. Grid computing supplied resource sharing across organisations. Utility billing supplied the pricing model. Seeing that lineage stops the field looking as if it appeared from nowhere.
Two properties are why anyone uses it. Scalability means a service can grow by adding machines, quickly, without a purchase order. Simplicity means someone else runs the physical layer, so a small team can operate something that would otherwise need a data centre.
Shared responsibility is the idea to carry forward. The provider secures the infrastructure. You secure what you put on it. Where exactly that line falls depends on which service model you are using, and misunderstanding it is how most cloud security failures happen. Nobody is guarding what each side assumed the other had.
The service level agreement is the part engineers skim and should not. It is the only place the provider states what it actually promises. It also defines what counts as an outage, and what you receive when the promise is missed.
Two things in it are worth reading carefully.
First, convert the availability figure into time before believing it. A promise of ninety-nine point nine per cent sounds close to perfect. It permits one part in a thousand of downtime, and one part in a thousand of a thirty-day month is about forty-three minutes. Every month, allowed, with no breach. Ninety-nine point ninety-nine per cent brings that to about four minutes. The gap between those two numbers is the gap between an inconvenience and a crisis, and it is one decimal place.
Second, the remedy is almost always a service credit rather than compensation. If an outage costs your business a great deal, the agreement typically returns a percentage of that month's fee. It does not cover your losses. Knowing that before depending on a provider is the practical value of reading it.
What you should now be able to explain or do
Explain why virtualization is what makes the economics work. Say what a hypervisor guarantees and what enforces that guarantee. Name the three technologies this field grew from. State the shared responsibility idea and say why it is where security failures occur. Convert an availability figure into minutes, and say what a service credit does and does not cover.
Check yourself
Why does virtualization make cloud computing economic?
It raises utilisation. Several customers' peaks fall at different times, so one machine serves many and the cost per customer falls.
What does the hypervisor guarantee, and what enforces it?
That one virtual machine cannot reach another's memory. It is enforced by software, which means it depends on that software being correct.
How much downtime does ninety-nine point nine per cent allow in a month?
About forty-three minutes. One part in a thousand of a thirty-day month, permitted with no breach of the agreement.
What is the usual remedy when a provider misses its target?
A service credit, which is a percentage of that month's fee. It does not cover the losses the outage caused you.
Why do most cloud security failures come from shared responsibility?
Each side assumes the other is covering something. Nothing guards the gap, and the boundary moves with the service model.
Go deeper
We haven't checked most of these for screen reader use yet.
Back to Virtualization & Service Level Agreements: work through the checklist