Edge Computing Cluster
for Real-Time Threat Detection

A 9-node Raspberry Pi cluster running real-time AI threat detection at 59 FPS via a Hailo AI accelerator, S3-compatible storage via MinIO, a k3s-orchestrated dashboard, MPI/HPL scalability benchmarking, and live Telegram alerting.
Cloud Computing SS2026 Prof. Dr. Christian Baun ✓ Tasks 1–9 Documented Presentation: 26 August 2026 github.com/Sapkale26/Cloud_Computing_Project
Documentation
System Architecture
Results
MPI & HPL Benchmarks
Live System
React Dashboard →
Key Performance Metrics
59 FPS
Hailo AI Inference
84×
Speedup vs CPU
9
Total Nodes
28
MPI Processes
12.99
GFlops (HPL Peak, Pi 5 alone)
Project Overview

We built a self-contained edge-computing cluster using commodity Raspberry Pis to demonstrate real-time AI-based threat detection and distributed cloud-computing concepts.

Key Capabilities
  • ⚡ AI Acceleration — Hailo AI HAT+ delivers 59 FPS compared with 0.7 FPS on CPU.
  • 🖥️ Distributed Computing — 9-node Raspberry Pi cluster for distributed workloads.
  • 🔄 Parallel Processing — 7 Raspberry Pi 3 workers handle distributed preprocessing.
  • ☸️ Container Orchestration — k3s Kubernetes manages workloads across the cluster.
  • 📊 Monitoring — Prometheus + Grafana provide cluster monitoring.
  • 📦 Object Storage — MinIO provides S3-compatible object storage.
  • 🤖 Alerting — Telegram Bot provides real-time threat notifications.
Cluster Architecture
Network Flow
Pi 4 (IMX500 Camera, 30fps)
    ↓ ZeroMQ PUSH
Pi 5 + Hailo AI HAT+ (59 FPS)
    ↓ every 5s: parallel preprocessing
Pi 3-1 ── Pi 3-2 ── Pi 3-3 ── Pi 3-4 ── Pi 3-5 ── Pi 3-6 ── Pi 3-7
    ↓ merged result
Backend API → MinIO → React Dashboard
                    → Telegram Alerts
                    → Grafana Monitoring
Hardware Nodes
Pi 5
Raspberry Pi 5 (8GB) · 192.168.50.1 · Master + Hailo AI HAT+
Pi 4
Raspberry Pi 4 (4GB) · 192.168.50.98 · Camera + ZeroMQ
Pi 3-1..7
Raspberry Pi 3B (1GB) · 192.168.50.91-97 · MPI + k3s Workers
Team
Shubhangi Sapkale
Janak Koradiya
Disha Bhuva
Kirti Tarsariya
Amina Arshad
Purvesh Shapariya
Marcos Ortega-Jimenez