Virtualization Showdown: Docker Containers vs VirtualBox VMs

Project Summary

This project conducts a comparative performance analysis between Docker containers and VirtualBox virtual machines on the same host system. The benchmarks used cover:

• Fibonacci Execution (CPU-bound load)
• Resource Elasticity Test (monitoring CPU/RAM usage under dynamic load)
• ML API Benchmark using Flask + scikit-learn, served via Gunicorn and load-tested using ApacheBench

Each test evaluates execution time, resource utilization, and system responsiveness under identical conditions to identify virtualization overheads and performance trade-offs.

How to Run the Benchmarks

1. Setup (Applies to both Docker and VirtualBox)

• Ensure Python 3.10+ is installed.
• Install the following Python packages:
  • scikit-learn
  • flask
  • gunicorn
  • psutil
• Clone the repository:

  git clone https://github.com/ashutosh0215/vcc-docker-vs-virtualbox-benchmark
  cd vcc-docker-vs-virtualbox-benchmark

• (For VirtualBox) Use shared folders or SCP to transfer files into the VM.

Docker Instructions

Build Docker Image

docker build -t vcc-benchmark:ml .

Run Fibonacci Benchmark

docker run -it vcc-benchmark:ml /usr/bin/time -v python3 benchmark.py

Run Resource Elasticity Test

docker run -it vcc-benchmark:ml /usr/bin/time -v python3 elasticity_test.py

Run ML API Server

docker run -it -p 5000:5000 vcc-benchmark:ml gunicorn -w 4 -b 0.0.0.0:5000 app:app

Load Test the API

ab -n 1000 -c 10 -p sample.json -T application/json http://localhost:5000/predict

VirtualBox VM Instructions

Run Fibonacci Benchmark

/usr/bin/time -v python3 benchmark.py

Run Resource Elasticity Test

/usr/bin/time -v python3 elasticity_test.py

Start ML API Server

gunicorn -w 4 -b 0.0.0.0:5000 app:app

Load Test the API (from a second terminal)

ab -n 1000 -c 10 -p sample.json -T application/json http://localhost:5000/predict