Multi-Tier Web Application Performance Simulation

This project implements a discrete-event simulation (DES) of a three-tier e-commerce web application using SimPy. It is designed as a full course project for studying capacity planning, performance, and cost optimization under variable load.

Features

• Three-tier architecture:
  • Web server pool
  • Application server pool
  • Database + cache layer
• Time-varying, non-homogeneous Poisson arrivals with daily pattern and load multipliers
• Detailed per-request timings across tiers
• Support for multiple scenarios:
  • Web/app server scaling
  • Cache hit ratio impact
  • Load multipliers (1x/2x/3x)
  • Combined “optimal configuration” search
• Multiple replications per scenario with random seeds and 95% confidence intervals
• Automatic storage of raw results and aggregated metrics
• Plotting utilities for all major figures required by the report

Test Configuration

For development and verification, a reduced test configuration is available:

• 10 req/s (vs 100 req/s baseline)
• 1 hour simulation (vs 24 hours)
• 3 replications (vs 10)
• ~10,800 total requests (vs ~8.64 million)

This runs in ~10 seconds and produces the same metrics as the full simulation.

Project Structure

project/
├── src/
│   ├── simulation.py        # Core SimPy model
│   ├── metrics.py           # Metrics collection and statistics
│   ├── config.py            # Baseline and scenario configurations
│   ├── experiments.py       # Scenario definitions and experiment runner
│   └── visualization.py     # Plotting and figure generation
├── data/
│   ├── raw_results/         # Raw per-run outputs
│   └── processed/           # Aggregated metrics
├── figures/                 # Generated plots for the report
├── notebooks/
│   └── analysis.ipynb       # Interactive analysis (optional)
├── requirements.txt         # Python dependencies
├── run_all_experiments.py   # Master script to run all scenarios
└── README.md

Quick Start

1. Create and activate a virtual environment (recommended).
2. Install requirements:

pip install -r requirements.txt

3. First, test the simulation (runs in ~10 seconds):

   python run_all_experiments.py --test

4. Run baseline configuration only:

   python run_all_experiments.py --baseline

5. Generate figures with real + synthetic data (recommended):

   python generate_figures.py --combined

   This uses your simulation results for available scenarios and fills in synthetic data for others.

6. Run all predefined experiments (~30+ minutes):

   python run_all_experiments.py

7. After runs complete:

   • Raw CSVs will be under data/raw_results/
   • Aggregated summaries under data/processed/
   • Figures under figures/

8. Open notebooks/analysis.ipynb in Jupyter for further exploration and to support the 15+ page report.

Notes for the Report

• The code is structured to map directly to the report sections:
  • Conceptual Model / Simulation Design → simulation.py, config.py
  • Data & Input Analysis → traffic and service-time utilities in config.py and plots in visualization.py
  • Experimentation & Results → experiments.py, run_all_experiments.py, metrics.py
• Use the generated CSVs and figures as the basis for tables, charts, and analysis described in the project prompt.