Foggy-sensors

Cloud & Fog Computing project — University of Trento
IaaS: OpenNebula · PaaS: Kubernetes (K3s) · Security: Network Policy + Falco Authors : Soranzo Andrea · Precoma Andrea

Overview

Simulated environmental sensors (temperature, smoke) send data via MQTT to edge fog nodes. Each fog node preprocesses the data locally and forwards aggregates to a central cloud cluster. Chaos Mesh injects node failures to test system resilience: when a fog node goes down, Kubernetes reschedules its workloads automatically. Optionally, a Python script calls the OpenNebula API to reprovision the failed VM and rejoin it to the cluster.

[Sensor A]  [Sensor B]  [Sensor C]
     \           |           /
      \        MQTT         /
       \        |          /
   [Fog Node 1] [Fog Node 2] [Fog Node 3 💥]
    VM+K3s       VM+K3s       VM+K3s (chaos target)
          \         |         /
           \   aggregated    /
            \    data       /
             [K8s Master VM]
          [Prometheus + Grafana]
               [Security]

Architecture

Layer	Technology	Role
IaaS	OpenNebula	Provisions and manages all VMs
Fog nodes	K8s on OpenNebula VMs	Local sensor data preprocessing
Cloud	Kubernetes master on OpenNebula VM	Cluster control plane, scheduling
Messaging	MQTT (Mosquitto)	Sensor → fog node communication
Monitoring	Prometheus + Grafana	Metrics, dashboards

Security Components

• Network Policy — namespaces isolated; only the aggregator pod may reach the cloud database
• Hardened containers — non-root user, read-only filesystem, securityContext in all manifests