Requirements to Join Edge Node¶

The edge nodes need to meet the specifications in the table below.

Operating System (OS)¶

x86_64 architecture

Ubuntu 16.04 LTS (Xenial Xerus), Ubuntu 18.04 LTS (Bionic Beaver), CentOS 7.x
armv7i (arm32) architecture

Raspbian GNU/Linux 9 (stretch)
aarch64 (arm64) architecture

Ubuntu 18.04.2 LTS (Bionic Beaver)

The general overhead of edge software is about 700MB, and it may increase appropriately depending on the pressure of edge management.

To ensure the normal operation of business, it is recommended that the memory of edge nodes should be more than 1GB.

Not less than 1 core.

Not less than 1GB.

Currently supports Nvidia Tesla series P4, P40, T4 and other GPU models.

Machines with GPU hardware can be used as edge nodes without GPUs.

If the edge node uses a GPU, you need to install the GPU driver and Nvidia Docker adaptation (nvidia-docker2) before registration.

You need to set the default runtime of the container engine to nvidia.
```
/etc/docker/daemon.json
```
```
{
  "default-runtime": "nvidia",
  "runtimes": {
    "nvidia": {
      "path": "nvidia-container-runtime",
      "runtimeArgs": []
    }
  }
}
```
Other requirements for NVIDIA GPU on ARM64 architecture:
- It is recommended that the system be installed with Jetpack4.5+.
- The image needs to be built using the base image hosted on NVIDIA L4T.
Environment verification:
```
nvidia-container-cli -k -d /dev/tty info
```
When the above command is successful, you can access it (The above tool comes with nvidia-docker2 after installation).

Docker version must be higher than 19.0, and it is recommended to use version 19.03.6.

Version must be higher than 2.17.

To check the version:

ubuntu: ldd --version

The edge nodes need to use the following ports, please make sure these ports can be used normally.

1883: The built-in MQTT broker listening port of the edge node, and this port needs to be opened.

The time of the edge node needs to be consistent with the time of the cloud server, and UTC standard time is recommended.

Version: >=3.6

To check the version:

python -V