November 18, 2025 - Docker Optimization for ML Models

🐳 Optimization Mission

Reduced Docker image size for machine learning model deployment by 40% while maintaining performance.

📦 Before Optimization

FROM python:3.9
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY . /app
WORKDIR /app
CMD ["python", "app.py"]

Image Size: 1.2GB

🚀 After Optimization

# Multi-stage build
FROM python:3.9-alpine as builder
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
 
# Final stage
FROM python:3.9-alpine
RUN addgroup -g 1000 -S appgroup && \
    adduser -u 1000 -S appuser -G appgroup
WORKDIR /app
COPY --from=builder /usr/local/lib/python3.9/site-packages /usr/local/lib/python3.9/site-packages
COPY --chown=appuser:appgroup . .
USER appuser
CMD ["python", "app.py"]

Image Size: 720MB

🔧 Key Optimizations

1. Multi-stage Builds

• Separate build and runtime environments
• Only copy necessary artifacts to final image

2. Base Image Selection

• Switched from full Python to Alpine
• Significant size reduction with trade-offs

3. Layer Optimization

• Order Dockerfile commands by frequency of change
• Combine RUN commands when possible

4. Security Improvements

• Non-root user execution
• Minimal attack surface

📊 Performance Metrics

⚠️ Trade-offs

Alpine Linux Considerations

# Some packages need special handling
# Problem: numpy installation on Alpine
RUN apk add --no-cache gcc musl-dev && \
    pip install numpy && \
    apk del gcc musl-dev

Binary Compatibility

• Some compiled packages behave differently
• Need thorough testing before production deployment

🎯 Best Practices Learned

1. .dockerignore

.git
.pytest_cache
.coverage
.venv
__pycache__
*.pyc

2. Health Checks

HEALTHCHECK --interval=30s --timeout=3s \
  CMD curl -f http://localhost:8000/health || exit 1

3. Resource Limits

# docker-compose.yml
services:
  ml-service:
    deploy:
      resources:
        limits:
          memory: 1G
          cpus: '0.5'

🔮 Future Improvements

Advanced Optimizations

• Distroless images for production
• BuildKit cache mounts for faster builds
• BentoML for ML-specific containers

Monitoring

• Prometheus metrics inside containers
• Resource usage tracking
• Automated image scanning for security

Mood: 😎 Proud of the results Containers Optimized: 3 Memory Saved: 480MB total