SovInfra / Continuity
Availability and Continuity
A multi-layer architecture for service resilience.
Hosting, orchestration, gateway and GPU capacity: each layer contributes to service continuity.
A data center foundation designed for 99.9999%
Our GPUs are hosted in European data centers designed for 99.9999% availability, with redundant active components, 2×N and N+2 configurations for power and cooling equipment, and a stated technical response time of under one minute, 24/7.
These specifications apply to the hosting infrastructure. On their own, they do not establish measured availability or an end-to-end contractual availability commitment for SovAPI.
Continuity at each layer
| Layer | Continuity arrangement |
|---|---|
| Orchestration | Primary server in France and standby replica in Falkenstein, Germany. |
| Access gateway | Standby server at a second site in another country. |
| Dedicated capacity | Shared capacity used during recovery, with the same models, region and endpoint. |
The recovery arrangement is intended to maintain service during a hardware incident. Failover procedures and the standby capacity applicable to your environment must be specified in the selected offer.
Measure. Report. Restore.
Service availability is assessed end to end.
Observed availability
Visit status.sovinfra.cloud for service status and incident history. This measurement complements data center specifications and helps assess actual service operation.
Performance and availability
The Arena measures latency and throughput for specific requests. Availability measures the ability of the service to respond over time. These indicators are complementary; a speed test alone does not establish an availability commitment.
Contractual commitments
Service-level agreements (SLAs), their measurement scope and any applicable penalties are defined in the contract. For dedicated capacity, the conditions for recovery on shared capacity must be agreed with the client.
A data center’s stated availability, observed API availability and the contractual SLA are three distinct measures.
Contact and resources
- Service status: status.sovinfra.cloud
- Performance measurements: arena.sovinfra.ai
- SovInfra team: contact@sovinfra.ai