Skip to content
Developper Doc
Requirements

Requirements

This section is a quick look at the software and hardware needed to integrate a target to the dAIEdge-VLab.

Tips: Please make sure to read this section carefully before starting the integration process. It will help you to understand the requirements and to avoid common pitfalls.

Hardware requirements

In order to integrate a new target you must have the followings :

  • Host : A PC. It may be a VM. It may be possible to use the target as the host PC.
  • Target : A board (OS-Based or MCU), or any system that can run AI models.
  • An Internet connection for the Host
  • A public access point to the Host (443 and 51820 ports must be open). This is necessary for the other Hosts of the cluster to communicate with your Host.
  • Necessary hardware to connect the Target to the PC (Ethernet dongle, USB cables, …)
Warning: The public access point requirement may be a problem for some users. Note that there is no going around this requirement as the Hosts of the cluster need to communicate with each other. The minimum requirement is to have the port 51820 open for the Wireguard communication.

Software requirements

  • Recommended OS for the Host: Ubuntu 22.04. You may use any other system that is supported by :

Knowing what you will implement

Before starting the integration process, it is important to have a clear idea of what services you want to provide with your target. The dAIEdge-VLab supports different types of services that can be implemented with the target. Each service has its own requirements and implementation complexity. It is important to choose the services that are relevant for your target and that you are able to implement.

The simplest service to implement is the benchmarking service with random data (TYPE1). It is a good starting point to get familiar with the integration process and the dAIEdge-VLab architecture. Once you have implemented this service, you can then consider implementing more complex services such as the benchmarking service with custom data (TYPE2) or the on-device training service (TYPE3). Be aware that these services are integrated in the same docker image.

Warning: Knowing what services you want to provide with your target is crucial to ensure a smooth integration process. Please take time to understand the different services.
Mandatory Guidelines