FAQ
Eight technical questions
The answers are factual and quantified. If your question is about a specific sizing, open a technical file: a pre-sales engineer replies within two working hours.
Bare metal: 112 seconds median through the API, for F-Series and D-Series profiles held in stock. Private cloud: 4 hours for a standard footprint, 5 to 15 working days if the sizing needs specific hardware. GPU nodes: within 48 hours on reserved capacity, 5 working days otherwise. Colocation: 10 working days, carrier cross-connect included.
Yes. Every action available in the console has an equivalent in the REST API, without exception: it is a design constraint, since the console consumes the same API you do. We maintain a Terraform provider and a CLI. Credentials are scoped tokens, individually revocable, with an exportable audit log.
Reversibility is contractual. You take your volumes away in standard open formats, your network configuration as readable descriptors, and your backups by direct export. No exit fee, no egress charge, thirty days’ notice. Our teams support the exit at the same service level as the onboarding.
RPO of 15 minutes: measured as the maximum gap between the last transaction committed on the primary site and the one present on the standby site, sampled continuously. RTO of 30 minutes: measured from the failover trigger to a nominal service, verified at the six-monthly test under real conditions. The record from the last test is available on request.
Four verifiable points. Data, backups and logs reside in our four French sites. Operations, on-call duty and the SOC are run from France by our own employees. The hypervisor, the orchestrator and the mitigation chain are built in house. No company under non-European law holds a stake in, or a right of access to, our infrastructure. We prefer those four statements to a single buzzword: each can be checked on its own.
Not on the sovereign private cloud: the CPU and memory allocation ratio is 1:1 and appears in the contract. Not applicable on bare metal, where the machine is physically dedicated. Guaranteed throughput per server is contractual and is not shared. On GPU nodes, reserved capacity is guaranteed and never reallocated to another queue.
Bare metal and private cloud: 12 months, then monthly renewal with 30 days’ notice. GPU: hourly with no commitment, or a quarterly reservation at a reduced rate. Colocation: 24 months, tied to the carrier cross-connect. Backup, disaster recovery and SOC follow the term of the service they protect.
Yes for colocation, and for bare metal on a reasoned request. Access requires 24 hours’ notice and goes through a single-person airlock with biometric control and photo identification. Every entry is logged and video is retained for 90 days. A site technician accompanies you throughout the visit.
A sizing question?
Generic answers stop here. For a specific case — volumes, latency, RPO — a pre-sales engineer replies within two working hours.