VPS sizing
These estimates assume ordinary mobile browsing at approximately 5-10 Mbps average per active device. They do not model simultaneous speed tests, sustained large downloads or video streaming on every device.
| VPS size | Adaptive profile | TCP flow ceiling | Fast devices | Balanced devices | Stealth devices |
|---|---|---|---|---|---|
| 1 vCPU / 1 GB | Small | 2,048 | 1-3 | 1-2 | 1-2 |
| 2 vCPU / 1 GB | Small | 2,048 | 2-5 | 2-4 | 1-3 |
| 2 vCPU / 2 GB | Standard | 4,096 | 3-8 | 2-6 | 2-5 |
| 4 vCPU / 4 GB | Performance | 8,192 | 8-18 | 5-15 | 4-12 |
| 4 vCPU / 8 GB | Performance | 8,192 | 18-45 | 15-40 | 12-30 |
| 8 vCPU / 16 GB | High capacity | 16,384 | 50-110 | 40-100 | 30-80 |
Interpret the flow ceiling correctly
Section titled “Interpret the flow ceiling correctly”A TCP flow is not a user or device. One modern phone can open tens or hundreds of concurrent flows. The ceiling protects the gateway from unbounded admission; it is intentionally much higher than the recommended active-device count.
Common bottlenecks
Section titled “Common bottlenecks”- Provider network throughput and fair-use policy.
- Xray CPU during encryption and high connection churn.
- Memory used by the web application, PostgreSQL and active flow state.
- Disk pressure from container logs, database growth and backups.
- Client network latency and packet loss.
The installer derives safe gateway defaults from available CPU and memory. It cannot infer provider bandwidth quality, CPU contention on a shared host or the intended workload.
Before promising capacity, run reproducible connection, throughput and restart
tests on the exact VPS plan. The full benchmark methodology remains available in
docs/SIZING.md.