CPU Utilization Forecasting with Federated Learning

This project studies the effectiveness of resolving the resource management (e.g., CPU) problem using the Federated Learning.

Install Dependencies

The only package you need in your operating system are Python3 and PIP3. The experiment was run on Python v3.9.6 and virtualenv==20.26.2

Getting Started

#### Starting a container

docker run --shm-size=4gb -it python:3.9 /bin/bash

### Install git and pop

apt -y update

apt -y install git

wget https://bootstrap.pypa.io/get-pip.py

### Install virtual Env

python3.9 get-pip.py

pip3.9 install virtualenv

### Retrieve the code from figshare

### Alternatively, the code can be retrieved from github

git clone https://github.com/fcrlab-unime/fl-cpu-utilization-prediction.git

cd fl-cpu-utilization-prediction

git checkout 04f17e7

### Start a virtual environment

python3.9 -m virtualenv -p python3 venv

source venv/bin/activate

### Install all Python3 dependencies

pip3.9 install -r requirements.txt

Datasets

The datasets are extracted from the AzurePublicDataset repository, a public collection of Microsoft Azure traces for the benefit of the research and academic community. Specifically, this repository uses the AzurePublicDatasetV2, which contains a representative subset of the first-party Azure Virtual Machine (VM) workload in one geographical region. Considering the dataset is very large with more than 2.5 million of VMs involved, we extracted a sample of 7 VMs.