Separate the management computer from the runtime node
The Windows 64-bit client adds servers, starts deployment, and handles routine management. The Ubuntu 24.04 amd64 node runs Docker, the node service, system images, and cloud phone instances.
The two devices can share a LAN or use a secure network path you configure, but the current Windows management computer cannot also act as the cloud phone runtime node.
Current deployment requirements
| Check | Current requirement | How to interpret it |
|---|---|---|
| Management OS | Windows 64-bit | Runs the client; it does not host the cloud phone runtime. |
| Node OS | Ubuntu 24.04 amd64 | The explicit support range for current one-click deployment. |
| Node memory | 4 GB minimum | A preflight threshold only; multiple phones require more memory and system headroom. |
| Node data storage | 30 GB free minimum | A preflight threshold; 80 GB or more is recommended for multiple images or instances. |
| Network and SSH | Client can reach the node | Requires the correct IP, port, account, password, and trusted host fingerprint. |
| Binder environment | Required | Deployment tries to load Binder and mount binderfs; follow the log if that fails. |
| KVM | Not a current hard requirement | Missing /dev/kvm produces a warning; operation can continue with limited virtualization capability. |
| Dedicated GPU | Not a current hard requirement | Without a detected GPU, the node uses CPU mode and must be validated under the target workload. |
Why one number cannot answer how many phones a computer can run
Each phone has CPU, memory, image, and network settings. The node also reserves resources for Ubuntu, Docker, the Agent, image cache, and startup peaks, so identical hardware can support different counts under different configurations.
The system considers node CPU, memory, load, and startup peaks when admitting work. A reliable approach is to validate one phone, add more gradually, and observe node metrics and startup results in the client.
- Processor: sustained multi-core performance matters more than model age alone.
- Memory: subtract system reserve before estimating per-phone allocations.
- Storage: system images, instance data, and logs all grow over time.
- Network: remote display, downloads, and concurrent instances share bandwidth.
- Cooling and power: especially important for older computers running continuously.
A five-minute pre-deployment checklist
- Back up important files before installing Ubuntu or changing disk partitions.
- Confirm Ubuntu 24.04 amd64 and correct system time on the node.
- Confirm SSH access and verify the host fingerprint from the management computer.
- Confirm memory, data storage, and network exceed the minimum with practical headroom.
- Prefer a direct LAN path when the client and node are on the same network.
- Run the client connection test and read-only preflight before deployment.
Frequently asked questions
There is no reliable count based on 4 GB alone. It is a preflight minimum, and the system and node service also use memory. Test the actual image, per-phone allocation, and workload.
It is only the minimum free space for deployment to continue. Images, instance data, and logs grow over time, so plan more storage and monitor it.
Neither is a current hard requirement. Missing KVM causes a warning, and a node without a detected GPU uses CPU mode. Validate the target workload on your hardware.
It depends on cooling, power, storage health, network stability, and sleep settings. The ability to install Ubuntu alone does not prove long-term multi-instance stability.
Free to use · Start with the tutorial
