Documentation: https://5uperpalo.github.io/success6g-edge/

Basic Architecture Overview, more detailed information in architecture_overview.md

This repository is for the edge pod implementation, monitoring, and analysis in SUCCESS6G project.

Table of contents

• detailed description of the use cases
• description of the data management and database choices
• description of the ml model choices
• explanation of the communication between the services
• description of Research&Development setup
• description of Testing setup
• inference model helm chart
• a guide to implementing needed services
• computational requiremetns of the services
• vehicle injector tool - tool to inject example v2x data into Redis database
• example v2x sensor data provided by Idneo
• example v2x aggregated sensor data provided by Idneo
• data and notebooks directories include analysis code used for initial edge model deployment and testing.

Solution overview

Solution is deployed in Microk8s.

Description of the components:

• Grafana - dashboards
• Ingress - expose services to the operator
• Prometheus - gather pod metrics
• InfluxDB - gather vehicular measurements and predictions
• MinIO - store models and training/testing data
• JupyterHub - develop new models
• MLflow - MLops, experiment and model tracking
• Kserve - serve inference models to predefined pods
• Istio - to ensure optimal traffic flow between microservices
• Knative - to ensure autoscaling of inference service pods
• Kepler - gather energy consumption data
• Redis - API for transfer of OBU measurements to Kubernetes

Additional ideas

• KubeEdge deployment - same as Microk8s except with KubeEdge and Kubeflow/Kserver is swapped for Sedna
• implement multimodel pods e.g. by ModelMesh, or alpha feature of Kserve
• use Rancher to manage multi cluster Kubernetes
• implement Kserve inference service as gRPC for high-performance/low latency production implementation