Virtualization & Microservice Architecture for Software-Defined Vehicles: An Evaluation and Exploration

1. Introduction

This repository includes codes employed in the paper Virtualization & Microservice Architecture for Software-Defined Vehicles: An Evaluation and Exploration.

NOTE: To properly use the deployment files for benchmarking, please replace all the paths like /home/tumi6 to your own path.

NOTE 2: We have modified the source code of ROS2. Please copy the folders in ROS2_source_code and replace them in the ROS2 folder.

NOTE 3: The container images are avaliable by searching longwen549 in the dockerhub.

NOTE 4: The provided container image is a full image that contains all required dependencies for the whole Autoware. However eliminating the unnecessary contents for different module images can further boost the performance and start-up time. For detailed information about the minimum setups for each module, please refer to eliminating.txt. These setups are tested by ourself and optimizations may still be available.

2. File Structure

• autoware_log: k3s log files and k3s start up time process scripts sorter.py, startup_time.py
• bash_folder: bash files to run test bare-metal-rosbag, k3s 17 pods, k3s 26 pods
• launch: launch files
• k3s_deployments: k3s deployment files
• test_result_process: process test results
• ROS2_source_code: source code for tracing launch of nodes

3. Installation

Autoware

Extract from the compressed archive and install dependencies

https://autowarefoundation.github.io/autoware-documentation/galactic/installation/autoware/source-installation/#source-installation

ROS2_galactic

ROS2 galactic need to be installed from source to ensure the launch of every node is reported. Debian install is also fine if you only want to have a try.

Important: The action folder in ROS2_source_code need to be pased to /path/to/ros/ros2_galactic/src/ros2/launch_ros/launch_ros/launch_ros and the rclcpp_components folder need to be pased to /path/to/ros/ros2_galactic/src/ros2/rclcpp/rclcpp_components/include and then recompile the whole ROS2.

To install dependencies of end-of-life distros

rosdep --include-eol-distros update

Docker

https://docs.docker.com/engine/install/ubuntu/

K3s (runtime Docker)

The default runtime of k3s is containerd, make sure using docker runtime when installing k3s.

https://docs.k3s.io/advanced#using-docker-as-the-container-runtime

NVIDIA Container Toolkit

Installation and configuration

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installation

After you install and configure the toolkit and install an NVIDIA GPU Driver, you can verify your installation by running a sample workload.

https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/sample-workload.html#running-a-sample-workload-with-docker

NVIDIA device plugin for Kubernetes

https://github.com/NVIDIA/k8s-device-plugin#enabling-gpu-support-in-kubernetes

4. K3s Deployment Workflow

Details in k3s_deployments\README.md

5. Data Processing Workflow

Details in test_result_process\README.md