The OVHcloud product range, decoded

OVHcloud‘s catalogue is broad and the naming is not always self-explanatory. Several products can host the same application, at quite different prices and with quite different operational characteristics.

Here is what each one actually is, and the question that decides between them.

The compute products

VPS is the entry point: a virtual private server on a fixed monthly plan, with a set amount of vCPU, memory and storage. Simple, cheap, predictable. Provisioned in minutes and administered by you. Suitable for a single application, a small website, a development box or a personal project.

Public Cloud Instances are the elastic option: virtual machines billed by the hour, created and destroyed through an API, with a range of instance families for general-purpose, CPU-optimised, memory-optimised and GPU workloads. This is the product that behaves like a conventional cloud, and it sits alongside a set of managed services such as object storage, managed databases and managed Kubernetes.

Dedicated Servers are physical machines on a monthly contract, in several ranges targeted at different priorities: value, infrastructure workloads, high-performance storage, and so on. You get the whole machine, and the price per unit of raw resource is far lower than virtualised equivalents.

Web Hosting is shared hosting with a control panel, aimed at conventional websites where you do not want a server at all.

The question that decides it

Ask in this order.

Do you want to administer a server? If no, and the workload is a conventional website, Web Hosting is the honest answer and it is much cheaper than the alternatives.

Does capacity need to change quickly, or be created by automation? If yes, Public Cloud Instances. Hourly billing and API provisioning are the whole point, and the fixed-plan products cannot do this.

Is the workload steady, predictable and large enough to use a whole machine? If yes, a Dedicated Server, where the price per core and per gigabyte is dramatically better.

Is it a single modest application with steady load? A VPS, which is the cheapest way to get a server you control.

Most confusion comes from comparing a VPS with a Public Cloud Instance on price alone. They are priced differently because they do different things: one is a fixed monthly plan, the other is elastic capacity billed by the hour. If you never scale, the VPS wins on price. If you scale, it cannot do the job at any price.

The managed services worth knowing about

Object Storage provides S3-compatible storage, which matters because compatibility means your existing tooling and libraries work unchanged, and it keeps the migration path open in both directions.

Managed Databases cover the common engines with backups, replication and patching handled. As elsewhere, the question is whether the operational saving justifies the premium over running the engine yourself, and for small teams it usually does.

Managed Kubernetes provides the control plane so you operate only the worker nodes. Worth it if you have already concluded you need Kubernetes; not a reason to conclude that.

Load Balancers and Private Networks connect the pieces. Private networking between your instances and dedicated servers is particularly worth setting up early, because it keeps internal traffic off the public internet and simplifies firewall rules considerably.

Bandwidth, which is unusually favourable

OVHcloud includes generous bandwidth with most products rather than metering egress aggressively. For bandwidth-heavy workloads such as media delivery, file distribution or anything with large responses, this is frequently the dominant factor in a cost comparison against providers that bill per gigabyte out.

If your current bill is dominated by data transfer rather than compute, this is worth modelling specifically, because the difference can exceed everything else combined.

The hybrid arrangement

As with any provider offering both physical and virtual compute, the strongest pattern is usually mixed: dedicated servers carrying steady baseline load, public cloud instances absorbing variable demand, joined over a private network.

Set the private network up at the start. Retrofitting it onto a running environment means re-addressing services and revisiting every firewall rule, which is tedious in proportion to how long you left it.

Practical notes

Provisioning times differ sharply. Public Cloud Instances appear in under a minute; VPS in a few minutes; dedicated servers from minutes to considerably longer depending on the range and current availability. Plan capacity ahead when dedicated hardware is involved.

Support and management levels differ by product, and the cheaper ranges are more self-service than the enterprise ones. Check what you are getting before assuming an incident will be handled the way it would be on a managed service.

For the vendor’s own reference on the services involved here, see the OVHcloud documentation.

Questions people ask