eco-bench/eco-bench

Setup for our edge container orchestrators (ECOs) comparison benchmarking

eco-bench

Setup for our edge container orchestrators (ECOs) comparison benchmarking

Planing

https://ecobench.atlassian.net/jira/software/c/projects/ECO/boards/1/roadmap?timeline=MONTHS

Directory Structure

.
├── .github/workflows               # Github Actions
├── docs                            # Documentation and notes
├── dummy                           # Smart plant factory dummy application
│   ├── camera-edge-device          # Camera
│   ├── image-processing-edge       # Image edge component
│   ├── image-processing-cloud      # Image cloud component
│   ├── sensor-edge-device          # Sensors
│   ├── sensor-edge                 # Sensor edge component
│   ├── sensor-cloud                # Sensor cloud component
│   ├── edge-irrigation-system      # Irrigation system
│   └── kubernetes                  # Kubernetes resources for the application
├── install-scripts                 # Scripts to automatically configure and provision ECOs
│   ├── kube-edge                   # Ansible script for kube-edge
│   ├── microk8s                    # Ansible script for microk8s
│   ├── openyurt                    # Ansible script for openyurt
│   └── provisioning                # Terraform script to provision VMs on Google Cloud
├── docker-compose                  # Docker-compose for the smart plant factory
└── benchmarking                    # Plotting and monitoring script

Installation

We have an install script automate_steps.sh to provision VMs, install mikrok8s, and deploy our use case application. The Script is located under install-scripts/microk8s.
The following steps are required to execute the script from you local machine:

1. Terraform and Ansilbe installed
2. service-account.json in provisioning folder, in google cloud console go to IAM & Verwaltung > Dienstkonten > Dienstkonto erstellen > Zugriff Standartt/Inhaber. After generating the Service Account click on the Service Account > Schlüssel and generate a new Key. Copy the key to the service-account.json file.
3. Create (or use an existing) Key Pair, and place it under Compute Engine > Metadaten > SSH-Schlüssel > bearbeiten.
4. In the ansible inventory file you will need your user name for the connection. Go to provisioning/inventory.tmpl and change it.
5. If your key file is not at the default location ~/.ssh/id_rsa you will need to provide the exact location and add this --key-file "~/.ssh/mykey.pem" whenever you run ansible-playbook in automate_steps.sh.

If you have completed all this steps execute the script from install-scripts/microk8s.

Forking

The project is designed to run the benchmarks through our pre-configured github action workflow which is defined in .github/workflows/terraform.yml

1. GCloud Setup

1. Enable the Compute Engine API: https://console.developers.google.com/apis/api/compute.googleapis.com/overview
2. Create your Service Account: https://console.cloud.google.com/iam-admin/serviceaccounts/
3. Create and download a JSON key for your Service Account (Actions > Manage Keys > Add Key)

2. ENV Preparation

Define the following secrets in Settings > Secrets > Repository secrets:

• PRIVATE_KEY
• PUBLIC_KEY
• GCE_SERVICE_ACCOUNT: The contents of your GCloud service account JSON file.

⚠️ Make sure the Service Account has sufficient access to your project. E.g. we have chosen the Owner role.

For now you also have to change some Terraform variables inside the files therefore open the install-scripts/provisioning/tfvars.json and edit the following values:

• gcp_project_id: The id of your google cloud project

3. Run the Action

To run the action in your own repository:

1. Go to the "Actions" tab
2. Select the "Install Edge Orchestrator" workflow on the left
3. Hit "Run workflow" on top of the list of runs

If you get an error regarding failed artifacts download, try to run the workflow again.

4. Destroy the Environment

The Install Edge Orchestrator workflow does not destroy the environment after a run to speed up the action on subsequent runs. If you need a clean environment use you have to destroy the previous environment with the Destroy Environment action. To run the action in your own repository:

1. Go to the "Actions" tab
2. Select the "Destroy Environment" workflow on the left
3. Hit "Run workflow" on top of the list of runs