Services

Six service lines

Each line is described by the operational benefit it delivers and by its exact technical scope. Pricing follows the sizing study.

SVC-01
SVC-02
SVC-03
SVC-04
SVC-05
SVC-06

SVC-01On study

Sovereign private cloud

A cloud of your own, operated at ours.

Dedicated hypervisor, dedicated hardware, dedicated network. You get a console and an API; we keep responsibility for the substrate. No resource is shared with another client, at any level of the stack.

Scope

Hypervisor built and maintained in house, on hardware reserved for you
Private L2 network stretched between data centres, VLAN and VXLAN under your control
Distributed NVMe storage, three synchronous replicas within the site
REST API, Terraform provider and CLI — every console action has a programmable equivalent
Granular RBAC, exportable audit log, quotas per project and per team
No CPU or memory overcommitment: the allocation ratio is contractual
Live migration between hypervisors, with no interruption of service
Machine-state backup included, retention configurable per project

overcommit_ratio: 1:1
replication: 3× synchronous
api: REST + Terraform

Request a sizing

SVC-02On study

High-frequency bare-metal servers

No hypervisor, no neighbour, no standard deviation.

For workloads that pay for every layer of abstraction: transactional databases, single-threaded compute engines, trading platforms. You get the bare machine, delivered by API in under two minutes.

Scope

Processors selected for frequency: up to 5.2 GHz under sustained load
DDR5 ECC memory, NVMe U.2 storage accessed directly, with no virtualisation layer
Throughput guaranteed by contract, from 1 to 10 Gbit/s depending on the profile — the throughput, not just the port speed
Delivered by API in under 120 seconds, with a stock or custom system image
Permanent IPMI and serial console; reboot and reinstall under your control
Hardware replaced within 2 hours, spares stocked in every suite

max_frequency: 5.2 GHz
provisioning: 112 s
guaranteed_throughput: 1 – 10 Gbit/s

Request a sizing

SVC-03Hourly or reserved, on study

GPU hosting and high-performance compute

Train and infer without queueing.

GPU nodes reserved or billed by the hour, interconnected at 200 Gbit/s and backed by a parallel file system. Capacity is guaranteed by reservation: you do not compete with other clients for your cards.

Scope

Compute nodes with 8 accelerators, high-bandwidth memory, liquid cooling
Inter-node interconnect at 200 Gbit/s, sub-microsecond RDMA latency
Parallel file system, 40 GB/s aggregate read across the cluster
Managed job scheduler, queues per team, quotas and priorities
Billed by the hour on shared capacity, or reserved quarterly with capacity guaranteed
Datasets held in France, with no transit outside the country

interconnect: 200 Gbit/s
parallel_fs_throughput: 40 GB/s
capacity: guaranteed by reservation

Request a sizing

SVC-04On study

Backup and disaster recovery

A recovery plan that has already been triggered for real.

Asynchronous replication between two data centres, immutable backups, and above all a failover test every six months, run for real and signed off in a written record. A recovery plan never triggered is not a recovery plan.

Scope

Contractual RPO of 15 minutes, contractual RTO of 30 minutes
Continuous asynchronous replication to a second data centre, 3.8 ms away
Backups immutable for 30 days: no deletion is possible, including by a compromised account
A full failover test every six months, with a signed and timed record
Granular restore, from a single file to the whole platform
A recovery runbook written with your teams and replayed at every test
Failover triggered by API or by call to the NOC, with dual approval

rpo: 15 min
rto: 30 min
immutability: 30 d

Request a sizing

SVC-05On study

Monitoring and security 24/7

The incident is detected before your first ticket.

NOC and SOC integrated, on our premises, in France. Event correlation, automatic DDoS mitigation, managed application firewall and incident response. You are told by an engineer, not by an automated alert.

Scope

NOC and SOC 24/7/365, in house — no on-call duty is subcontracted
SIEM correlation, with detection rules written and maintained by our R&D team
Automatic DDoS mitigation: 6 Tbit/s of capacity, 30 s from detection to neutralisation
Managed application firewall, rule sets adapted to your stack
Incident response: taken up within 15 minutes on severity 1
Monthly security report and a quarterly review with your CISO

mitigation_capacity: 6 Tbit/s
sev1_pickup: 15 min
subcontracting: none

Request a sizing

SVC-06On study, by density

Colocation

Your hardware, our fortress.

You already own your machines and want them back in a site that is properly powered, properly connected and properly guarded. Private racks or half racks, from 3 to 20 kW, across all four data centres.

Scope

42U and 47U racks, private or shared. Standard density 3 to 8 kW, up to 20 kW in a contained aisle
2N redundant power, dual feed, UPS and generators tested monthly
Multi-carrier transit, public and private peering, L2 transport between our sites
Smart hands 24/7, on site within 30 minutes, with time-stamped photographs
Biometric physical access, single-person airlock, video retained for 90 days

max_density: 20 kW / rack
power: 2N
smart_hands: 30 min

Request a sizing

We do not publish prices. What an architecture costs depends on the load profile, the guaranteed throughput, the power density and the recovery level chosen — four variables no price list can summarise. The costing arrives with the diagram, never before it.

Sizing

None of these services is ordered online

Every architecture is sized against your real load: volumes, read and write profile, latency constraints, target RPO and RTO. Send those, get a diagram and a costing back.

Open a technical file