Plan a private mobile device network
Scope a dedicated mobile proxy fleet from measured workload demand. Define session isolation, recovery, monitoring and commercial requirements before scaling.
Start with a representative workload on a small number of ports. Measure request demand and session behavior, then specify the countries, carriers, isolation and recovery capacity the fleet needs. Fleet size alone does not establish an uptime guarantee.
Write the workload specification first
A fleet supports a particular application, so capacity should follow that application's behavior. Record the number of independent sessions, the request mix, peak demand and the time a task keeps its connection occupied. Long browser sessions and short HTTP requests can require very different allocations.
Specify whether tasks may share a modem, whether they must retain a browser context and what happens when the carrier changes the public IP. A dedicated physical modem does not itself guarantee a permanently fixed address.
| Requirement | What to specify |
|---|---|
| Geography | Required countries and any carrier requirements; confirm catalogue availability. |
| Session isolation | Which workflows must have independent ports or browser contexts. |
| Demand | Measured request rate, response sizes, burst patterns and session durations. |
| Recovery | Allowed interruption, maximum retry time and task restart behavior. |
| Operations | Monitoring owner, escalation path, log retention and credential handling. |
Size from a measured trial
Run a controlled test using an authorized destination and a realistic request mix. Increase concurrency gradually while recording latency, failures and useful throughput. Stop increasing it when the application becomes unreliable; successful socket creation is not a throughput target.
Calculate a working port estimate from the measured capacity, then check whether session isolation requires more ports. Keep headroom for your expected bursts and recovery needs. The capacity guide and calculator provide explicit formulas without assuming an unlimited number of connections per device.
Any numbers in your planning model are assumptions until you measure them. A plan containing ten or one hundred devices does not, by itself, justify a 99.9% or higher availability claim.
Test the recovery path
A spare port helps only when the scheduler and application can use it. Tasks with stateful sessions may need to restart, and a timeout does not prove that a state-changing request failed before reaching its destination. Define those outcomes before promising recovery behavior.
- Stop assigning work to one port in a controlled test.
- Check that the scheduler marks it unavailable and stops sending new work.
- Let active tasks finish, cancel or recover according to their state.
- Use a confirmed spare connection only where the task can tolerate changing network state.
- Verify the recovered port before returning it to the pool.
Observe connectivity and application outcomes separately
Use small health checks with timeouts, bounded retries and a clear escalation threshold. Track connection failures separately from destination rejections, since an available proxy can still receive a 403 or 429 from a target service.
Keep credentials out of monitoring messages and logs. Record port references, timestamps and error categories. Retain detailed response data only when there is a specific diagnostic need and access is restricted.
Confirm the commercial scope
Before ordering a fleet, send support the required countries, carriers, port count, rental period and representative workload. Confirm current stock, any management requirements and the written service terms that apply to your order.
Custom pricing, service-level commitments and recovery arrangements need explicit agreement. The public rental prices are useful for an initial estimate; they are not a quote for a managed enterprise deployment.
- Required locations and number of independent sessions.
- Expected traffic, peak demand and test results.
- Credential ownership and access requirements.
- Monitoring and support expectations.
- Requested start date and rental duration.