This repository provides a comprehensive benchmarking framework for analyzing Node.js application performance patterns in Kubernetes environments. The project compares different Node.js server architectures under various load conditions and resource constraints to provide data-driven insights for production deployment decisions.
This benchmarking suite is designed to:
- Compare Node.js server architectures in realistic Kubernetes deployment scenarios
- Analyze performance characteristics across different CPU/memory resource allocations
- Evaluate scaling patterns under mixed workload conditions
- Provide production deployment recommendations based on empirical data
- Demonstrate Kubernetes resource management impact on Node.js applications
The project benchmarks three distinct Node.js server architectures:
- Single Process Server - Traditional single-threaded Node.js application
- Worker Thread Server - Node.js application utilizing worker threads for CPU-intensive tasks
- Cluster Server - Multi-process Node.js application using the cluster module
Each server configuration exposes two endpoints representing different computational loads:
/light- Light CPU-bound task (target: <50ms response time)/heavy- Heavy CPU-bound task (target: >300ms response time)
The suite tests performance across 10 different load distribution patterns to simulate real-world traffic scenarios:
- 10% light / 90% heavy - Heavy computation dominant workload
- 20% light / 80% heavy - High computation workload
- 30% light / 70% heavy - Computation-heavy workload
- 40% light / 60% heavy - Moderate computation workload
- 50% light / 50% heavy - Balanced workload
- 60% light / 40% heavy - Light computation workload
- 70% light / 30% heavy - Response-time optimized workload
- 80% light / 20% heavy - Fast response workload
- 90% light / 10% heavy - Low latency workload
- 100% light / 0% heavy - Pure light request workload
The benchmarks test multiple Kubernetes resource allocation scenarios:
- CPU: 500m, 1000m, 1500m, 2500m millicores
- Memory: 256Mi, 512Mi, 1Gi
- Workers: 1 (single), 4-10 (cluster/worker configurations)
Based on 120 comprehensive benchmark runs:
- Cluster configuration provides best overall performance and reliability (78.38% success rate)
- 100% light request workloads achieve 3x better performance than mixed loads
- Linear scaling observed in cluster configurations with increased resources
- Worker thread configurations show reliability issues requiring investigation
- Resource efficiency varies significantly between server architectures
- Build & Run Guide - Setup and local development instructions
- Kubernetes Deployment - Container deployment and K8s configuration
- Benchmarking Guide - How to run automated performance tests
- Analysis Report - Comprehensive performance analysis and findings
- Performance Expectations - Hypothesis and expected behavior patterns
The benchmarking suite has analyzed 120 test configurations and provides:
- Performance comparison across server architectures
- Resource efficiency analysis
- Production deployment recommendations
- Visual performance graphs and scaling patterns
- express - Web application framework
- ws - WebSocket implementation for real-time benchmark reporting
- autocannon - HTTP/1.1 benchmarking tool
- inquirer - Interactive command line interface
- pandas - Data analysis and manipulation (Python)
- matplotlib - Plotting and visualization (Python)