Cloud Hosting: What It Really Means (and What to Check Before You Choose)

"Cloud hosting" is one of the most used phrases in the industry, and one of the least precise. For some hosts it is a VPS with a new name. For others it is a platform where you create machines on demand and pay for what you use. For a buyer the consequence is simple: the word "cloud" tells you nothing about what you will actually get. Here is how to read it.

What "cloud" means, strictly speaking

The reference definition is the one from NIST, the US standards institute, published in 2011 (NIST SP 800-145). It lists five characteristics:

1. On-demand self-service. You create or delete resources yourself, without asking a person at the provider.
2. Broad network access. Resources are driven remotely, through a web interface or an API.
3. Resource pooling. The provider shares its machines among several customers, on a multi-tenant model.
4. Rapid elasticity. Capacity can be added or removed quickly, sometimes automatically, following demand.
5. Measured service. Usage is measured, controlled and reported, which makes it possible to bill and to follow.

So it is not a product, it is a way of delivering infrastructure. Hosting that ticks all five boxes is cloud in the strict sense. A server rented by the month and set up by a technician after an e-mail request is not, even if the brochure says otherwise.

Cloud, shared, VPS, dedicated: not exclusive categories

The confusion comes from comparing things of different kinds.

• Shared, VPS and dedicated describe what you receive: a share of a shared server, an isolated virtual machine, a physical machine for you alone. Our articles VPS or shared hosting and dedicated server or VPS go through those differences.
• Cloud describes how it is delivered and billed: on demand, measured, quick to change.

A VPS can therefore be cloud or not. What matters is what you can do without going through a human: create a machine, change its size, delete it, take a snapshot, see the usage.

The two ways cloud is billed

Without quoting figures, you need to tell two models apart, and check which one applies before you sign.

• The flat plan. A fixed configuration (vCPU, memory, disk) for a fixed monthly price. The bill is predictable. It is the simplest choice for a site or application with a steady load.
• Usage-based billing. Each reserved resource is counted, often by the hour: processors, memory, disk, IP addresses. It suits loads that vary a lot (a test environment switched on a few hours a week, a seasonal peak), but it takes discipline.

In both cases, ask these questions, which are the ones that cause bad surprises:

• Is a stopped machine billed? With many providers it is, because its resources stay reserved for it. To stop paying, you have to delete it.
• Is outgoing traffic counted? And from what volume?
• Are snapshots, backups and extra IP addresses included or billed separately?
• What happens when you change size? Is a restart needed? Can you go back?
• Can the bill run away? Is there a cap, an alert?

To understand what weighs on a hosting bill in general, read what hosting costs.

Elasticity: rarer and less automatic than you might think

There are two ways to add capacity.

Vertical means making one machine bigger (more vCPU, more memory). It is simple, but it has a limit, the largest configuration on offer, and it often needs a restart.

Horizontal means adding machines behind a load balancer. It goes much further, but it requires the application to run on several servers: shared sessions, files on common storage (an S3 object storage often serves there), a separate database. A classic WordPress site is not built for this without work.

In other words, cloud makes elasticity possible. It does not make it free. If your application fits on one machine, being able to go to twenty machines gives you nothing.

Cloud does not mean always available

Another common confusion: cloud would mean high availability. It does not. A single virtual machine on a cloud platform can go down, like any server. What changes availability is architecture: replicated storage, redundant servers, load balancing, tested backups. Our article on high availability explains what to put in place and what it costs in complexity. And the availability commitment is read in the contract: see how to read an SLA.

Finally, a cloud does not replace your backups. A machine deleted by mistake is gone for good. The 3-2-1 strategy applies here as anywhere.

When cloud is the right choice, and when it is less so

Cloud brings the most in these situations:

• a load that varies (launch, sales, season);
• test and staging environments you create and delete;
• a project whose size you do not know yet;
• a need to rebuild a machine quickly and identically.

It brings little, or complicates things, in these:

• a brochure site or a stable blog: web hosting is often enough;
• a heavy, constant load: a flat-plan VPS or a dedicated server is more predictable;
• a team without systems skills: better to delegate administration, see our article on the managed VPS.

At Atlas Hoster

Our Cloud VPS are KVM virtual servers on Proxmox clusters, with replicated NVMe storage, verified backups and continuous monitoring. You keep root access or hand administration over to our managed services team. Each offer shows its configuration (vCPU, memory, disk) and its billing on its own page, with a choice of payment frequency. If you are not sure what size to take, the article sizing your VPS gives the method, and you can ask for a quote for a particular need.

Frequently asked questions

Is a VPS cloud?

It depends on how it is delivered. A VPS you create, resize and delete yourself from an interface, with visible usage, meets the NIST criteria. A VPS delivered after a manual request and then frozen meets them much less.

Is cloud more or less expensive than a classic server?

Neither, as a rule. It is often cheaper for a varying load, and more expensive for a constant, high load. Compare by adding up everything that is billed (machine, disk, traffic, backups, IPs), not only the price of the machine.

Does cloud guarantee my site will never go down?

No. A single machine can go down. Availability comes from architecture (redundancy, replicated storage, load balancing) and from the written commitment in the contract, not from the word "cloud".

Is my data safer in the cloud?

Not automatically. The provider protects the infrastructure; you remain responsible for the configuration of your machines, their updates, your access and your backups.

Sources

• Mell P., Grance T., The NIST Definition of Cloud Computing, NIST Special Publication 800-145, September 2011.
• Features of our VPS: Cloud VPS page.