Newer
Older
# Software-Defined Industrial System (SDIS)
## Project Overview
SDIS is a reference implementation that proves the viability of using cloud-native approaches and virtualization in resource-constrained embedded environments for industrial automation and energy management. The system supports concurrent execution of near real-time control and compute-intensive workloads like machine vision/AI, all while maintaining levels of deterministic performance.
## Key Features
- **Hardware-Independent Design**: Built for Arm SystemReady certified platforms, enabling easy portability across different hardware
- **Real-Time Performance**: Supports deterministic operation for critical workloads
- **Mixed-Criticality Support**: Successfully runs both real-time control and best-effort AI workloads concurrently
- **Virtualization**: Optionally uses XEN hypervisor for secure workload isolation and resource management
- **Flexible Deployment**: Supports dynamic workload combinations (1 real-time + up to 4 best-effort workloads)
- **Edge AI Capable**: Demonstrated with multi-camera machine vision processing using Arm NN
## Technical Stack
- **Hardware**: Arm SystemReady certified platforms (tested on Xilinx ZCU102 and Raspberry Pi 4)
- **Hypervisor**: XEN (bare-metal configuration)
- **Real-Time Support**: PREEMPT-RT Linux patch
- **AI/ML**: Arm NN inference engine with GPU acceleration
- **Containerization**: Optional container support for both workload types
## Benefits
- Reduced development time through hardware abstraction
- Lower latency for critical applications
- Enhanced security through workload isolation
- Decreased reliance on cloud resources
- Improved sustainability through efficient resource utilization
## Getting Started
### Bluegreen machines
This project contains [kas](https://kas.readthedocs.io/en/latest/index.html) configuration to download meta-bluegreen and all required layers.
[EWAOL](https://gitlab.arm.com/ewaol/meta-ewaol) is the reference implementation of the [SOAFEE](http://soafee.io) project and is used as a baseline for bluegreen
project.
You must install essential host packages on your build host. The following command installs the host packages based on an Ubuntu distribution:
```shell
sudo apt install gawk wget git diffstat unzip texinfo gcc build-essential chrpath socat cpio python3 python3-pip python3-pexpect xz-utils debianutils iputils-ping python3-git python3-jinja2 libgl1 libglx-mesa0 pylint xterm python3-subunit zstd liblz4-tool
```
Install the kas tool, ensuring you have a version 3.0.2 or greater. The preferred method is installation using the package manager of your Linux distribuntion.
#### 1.a. Preferred: Install kas using package manager on Linux
On Ubuntu Linux:
```shell
sudo apt install kas
```
If you can't install kas as above, you can follow the instructions [here](https://kas.readthedocs.io/en/latest/userguide.html).
However, please note some issues where encountered when using kas installed from pip. See Known issues, below.
To build bluegreen, use the helper script `build.sh` which gives a consistent way of building for the target platforms.
`build.sh` is located in the root of this project. It relies on kas to do the actual build.
```shell
# Building for the generic-arm64 target
./build.sh generic-arm64 baremetal
```
Installation instructions for each platform is outlined in the documentation for that platform. Check out the table in the status section below.
| Machine | Status | Documentation |
|---------------|--------|---------------|
| generic-arm64 | Ok | [Generic arm64 Documentation](docs/generic-arm64.md) |
Put UEFI firmware on the sdcard and the bluegreen image on a SD Card.
See [ZCU102 SD card boot documentation](docs/zcu102_sd_boot.md).
This has been tested on the RPi4 using [SystemReady compliant UEFI firmware](https://github.com/pftf/RPi4).
* On Ubuntu 22.04, using `pip` to install `kas` in a Python virtual environement produced a tool that wasn't able to finish the build. Root cause is unknown, workaround is to get `kas` from with `apt install`:
```shell
sudo apt install kas
```
* On Ubuntu 22.04, it was necessary to manually install libtinfo5:
```shell
sudo apt install libtinfo5
```