# System Requirements - Mission Systems ## Machines Four Linux machines, all cabled to the same Moxa Time-Sensitive Networking (TSN) switch and to a lab network: | Machine | Role | Switch port | | --------- | --------------------------------------------------------------------------------------------------------------- | ----------- | | machine-1 | Dashboard, Scenescape, Deep Learning Streamer (DL Streamer), MQTT broker. Also the machine where you run `make` | 1 | | machine-2 | Camera 1: test-video stream and inference | 2 | | machine-3 | Camera 2: test-video stream and inference | 3 | | machine-4 | Load injector | 4 | Each machine needs an Intel I225-LM or I226-LM TSN network card (PCI ID `8086:15f2` / `8086:125b`). Machine-1 needs about 8 CPU cores for the full stack and several GB of free disk space for the Docker images (the DL Streamer image alone is about 5 GB). The camera machines need Docker and enough CPU for one inference pipeline each. ## Network Topology ![Network topology: four machines, the Moxa TSN switch, optional Basler cameras, and the lab network](../deterministic-threat-detection/_assets/network-topology.svg) Ports 1 to 4 carry VLAN 1 (camera data, priority 1), VLAN 3 (Basler cameras), and VLAN 5 (load, priority 5). The dotted lab network is never shaped. ## Software - Ubuntu 24.04 - Docker Engine with Docker Compose V2 on all machines, so that `docker compose` works (not only the legacy `docker-compose`). For installation instructions, see the [installation guide](https://docs.docker.com/engine/install/ubuntu/). - `git` and `make` on machine-1 - `linuxptp` and NetworkManager (installed or used by the setup scripts) - Passwordless, key-based SSH from machine-1 to machines 2, 3, and 4 - Internet or proxy access on machine-1 to pull the Docker images and to download the detection model. The download script (`src/data/download_models.sh`) runs automatically in `make up`. The model comes from Open Model Zoo (Apache-2.0), and `make up` asks you to accept its license on the first download. ## Network - **Lab network:** SSH, the dashboard in the browser, and control traffic. Not part of the TSN path. - **TSN switch:** VLAN 1 (`192.168.127.x`) camera data, VLAN 5 (`192.168.5.x`) background load, VLAN 3 (`192.168.3.x`) Basler cameras. ## Hardware | Component | Details | | ---------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- | | Moxa TSN-G5000 switch | IEEE 802.1Qbv and IEEE 1588v2/802.1AS support, Telnet management enabled (the dashboard scripts use it). Tested with TSN-G5008-2GTXSFP, firmware 2.1 | | Basler ace GigE cameras (optional) | Precision Time Protocol (PTP) hardware timestamps, for the Basler test only. Tested with acA1920-40GC |