A simulation-based system that optimizes cloud workload execution by shifting tasks to time periods with lower carbon intensity, while balancing delay and cost constraints.

This project implements a carbon-aware scheduling engine that evaluates when to execute workloads based on environmental impact. Instead of running jobs immediately, the system intelligently selects future time slots that minimize carbon emissions without introducing excessive delay.

This phase focuses on generating a realistic dataset to simulate cloud workloads and carbon intensity patterns.

• Hourly timestamps over a 90-day period
• Simulated workload demand with daily and weekly trends
• Carbon intensity variation (gCO₂/kWh)
• Time-series structure suitable for scheduling decisions

data/carbon_aware_workload_dataset.csv

The project is built around three main components:

• Naive Scheduler: Executes jobs immediately

• Carbon-Aware Scheduler: Selects optimal execution time within a 24-hour window

• Multi-objective scoring based on:

  • Carbon intensity
  • Execution delay
  • Cost (simplified)

• Runs both naive and smart scheduling strategies

• Computes:

  • Total carbon emissions
  • Carbon savings (kg)
  • Percentage reduction
  • Average delay per job

• Built using Streamlit

• Real-time tuning using sliders:

  • Carbon weight
  • Delay weight
  • Cost weight

• Visualizations:

  • Total carbon comparison
  • Carbon intensity over time
  • Per-job carbon comparison
  • Delay distribution

• Achieves measurable carbon reduction (~10–20%) under realistic constraints
• Maintains low average delay per job
• Demonstrates trade-offs between sustainability and performance

For each job:

1. Identify candidate execution times within the next 24 hours

2. Compute a score for each candidate:

   score = w_carbon * carbon + w_delay * delay + w_cost * cost
   

3. Select the time with the minimum score

4. Compare against naive scheduling

• Python
• Pandas
• Streamlit
• Plotly / Matplotlib

pip install -r requirements.txt
streamlit run src/dashboard/app.py

• Integration with real-world carbon intensity APIs (Electricity Maps, WattTime)
• Job-specific deadlines and SLA constraints
• Multi-region scheduling
• Dynamic cost modeling
• Advanced optimization techniques

This project demonstrates how intelligent scheduling can reduce carbon emissions in cloud systems by leveraging time-based optimization. It highlights the importance of balancing environmental impact with system performance in modern cloud computing.