ICOS FL: Federated Learning for Resource Monitoring

ICOS-FL is a federated learning framework powered by Flower for real-time resource monitoring and prediction. It enables distributed training of LSTM models to forecast CPU usage, memory consumption, and power consumption across ICOS nodes while preserving data privacy.

Features

• Federated Learning: Train models across distributed nodes while keeping data local
• Privacy-Preserving: Learn from system metrics without centralizing sensitive data
• Real-time Monitoring: Track CPU, memory, and power consumption metrics
• LSTM Prediction: Forecast resource usage with configurable time windows
• DataClay Integration: Efficient storage and retrieval of time series data
• Docker Deployment: Easy setup with containerized components
• Modular Architecture: Customize models, metrics, and federated strategies

Requirements

• Python 3.10 or newer
• Docker and Docker Compose
• At least 4GB of available RAM
• Network access between nodes (for distributed deployment)

Installation

From Source

bash git clone https://github.com/anaskalt/icos-fl.git cd icos-fl pip install -e .

Quick Start

Setup Infrastructure

```bash

Start required services using Docker Compose

docker compose -f docker/simulation.yml up -d ```

Run Federated Learning

```bash

Start FL

flwr run . local-deployment --stream

Start FL with different configuration

flwr run . --run-config "metric=power_consumption" --stream ```

Architecture

ICOS-FL consists of three main layers:

1. Data Collection Layer: Captures system metrics via Scaphandre and OpenTelemetry
2. Storage Layer: Persists time series data in DataClay with sliding window approach
3. Learning Layer: Implements federated learning with Flower's SuperLink and SuperNodes

The framework uses LSTM (Long Short-Term Memory) neural networks to predict resource usage patterns based on historical system metrics.

Documentation

Comprehensive documentation is available at https://icos-fl.readthedocs.io/, including:

• Introduction to ICOS-FL
• Architecture Overview
• Deployment Guide
• API Reference

Contributing

Contributions are welcome! Please check out our Contributing Guide for details on:

• Code Guidelines
• Documentation Standards
• Testing Requirements

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

If you use ICOS-FL in your research, please cite:

bibtex @software{kaltakis2025icos, author = {Kaltakis, Anastasios}, title = {ICOS-FL: Federated Learning Framework for Resource Monitoring}, year = {2025}, url = {https://github.com/anaskalt/icos-fl} }