Documentation
Sizing
Sizing a deployment means starting from what is measured, then applying the gate objectives and the reference vigie template. This page separates the two: throughput figures come from a local bench and are stated with their condition; the gate objectives and the template are design targets, not measurements.
MeasuredMeasured throughput
Our throughput figures come from a local multi-process bench: distinct OS processes exchanging over real sockets, on loopback and in-memory TUN. They frame the order of magnitude of one processing core, not the behaviour of a real link.
Relay gate objectives
A relay is only retained if it sustains a per-core service threshold. This threshold acts as a gate: below this level, the core is not considered fit to relay production traffic. It is a design target, not a result already measured in production.
Sizing a relay therefore comes down to multiplying this per-core budget: add up cores until you cover the expected peak throughput, with a margin for spikes and supervision.
Reference vigie template
For the vigie, we start from a reference template that serves as the starting point for the calculation. It expresses a design order of magnitude, to be adjusted to the actual observed load.
| Resource | Reference template |
|---|---|
| Processor | 4 vCPU |
| Memory | 8 GiB |
| Network throughput | 5 Gbps |
| Targeted concurrent connections | about 10,000 |