A build matched to the load
High clock speed where databases are billed per core, memory generous enough to stop the system going back to disk, and NVMe anywhere the count of operations per second outweighs raw capacity.
A dedicated server earns its place when sharing resources starts costing you performance. The database behind the business system passes a few hundred gigabytes. Overnight jobs spill into the morning. Or a major customer writes into the contract that your data must not sit on the same hardware as anybody else. At that point you rent a whole physical machine from the provider. The hardware stays theirs: they power it, cool it and swap failed parts. We take responsibility for everything running on top. You buy no equipment, and nobody travels anywhere, neither you nor us.
What bare metal really buys is predictability: a build that depends on no neighbour, figures that hold steady from one month to the next, and a split of responsibility written down between the provider and us.
High clock speed where databases are billed per core, memory generous enough to stop the system going back to disk, and NVMe anywhere the count of operations per second outweighs raw capacity.
Region, redundant power and network paths, a stated time to replace failed parts, and processing terms clear enough to record in your GDPR register. An HDS-certified provider as soon as health data is involved.
Disk array, operating system, networking, filtering, and above all the provider out-of-band console, our way back in when the system has stopped answering altogether.
Some databases and business modules are priced per core. Many slow cores can therefore attract a licence bill well above that of fewer fast ones, while performing worse. Hardware and licences are costed together for that reason.
Backups written to separate storage and, for operations that cannot stop, a second machine kept in replication and ready to take over.
Updates, watching disk and array health, raising and chasing tickets when hardware fails, plus a yearly look at whether the specification still suits the business.
Lead time depends on the build chosen and on its availability at the provider. The database move is always rehearsed before the real cut-over.
Across one representative week we log processor, memory and disk activity, the monthly close included, since that peak is what sets the specification.
The machine is ordered in your company name, then we build the array, the system, the filtering and the monitoring before any data goes near it.
A rehearsal migration measures how long the operation really takes and flushes out the surprises. The real cut-over then happens outside production hours.
Monitoring, patching and configuration adjustments as the volume of data and the number of connected users grow.
A dedicated machine is not a backed-up machine. The confusion comes up often: because the server belongs to you alone and runs a RAID array, people assume the data is safe. An array absorbs the failure of one disk and nothing more. It does not undo a careless deletion, does not rescue a failed update, and offers no resistance to ransomware, which will encrypt both disks with equal enthusiasm. Copies therefore go elsewhere, to a different provider, and we restore from them for real at regular intervals.
Three signals keep coming up. A closing report or stock listing that takes ten minutes to appear. Slowdowns at the same hours every day, despite two successive increases in plan size. A contract clause that rules out shared hardware. If none of the three applies to you, stay on a VPS: it is cheaper and it resizes far faster.
Monitoring warns us before the array loses its redundancy. We raise the ticket, a technician on site swaps the part, and we then verify that the array rebuilds and the data has come through intact. No spares are bought by you, and nobody from your side travels to the facility; nor do we, since the whole thing is driven remotely.
Yes. The providers we work with offer French regions alongside other European locations. We pick a French region when a customer, a public buyer or your own internal policy requires it, and that location is recorded in the documentation handed to your data protection officer.
We sell no hardware, so neither answer suits us better than the other. Buying stands up where the load is steady and the technical room is already cooled and on a protected supply. The sum then has to include the UPS and its batteries, the maintenance contract, the floor space tied up, and the person who will end up swapping a disk one Saturday. Renting wipes those lines out and replaces them with one monthly invoice.
The split is written into the contract. The provider answers for the building, the electricity, the network and the hardware. We answer for the system, the applications, the backups and the monitoring. We are the ones who raise the ticket, chase it and keep you informed, Mon-Fri 8:00-18:00 CET, with a response time set by your plan.
Set out the load, how big your databases are and any constraints on where data may sit. We propose a build, the licence costs that go with it and a comparison against an equivalent VPS.
Message received
An answer follows inside one working day. Report an outage that is stopping people working and it moves ahead of everything else.
Nothing here under that name. Check the spelling, or simply choose the nearest large city instead. Since every engagement runs remotely, whichever you pick changes nothing about what we do for you.