Backend Framework Performance Comparison

Load testing comparison of multiple backend frameworks/languages using k6.

Test Results Summary

Rank	Framework	Throughput	Avg Response	Runtime
🥇	Bun Native	30,733 req/s	32ms	Bun (Zig)
🥈	Go	28,597 req/s	35ms	Go
🥉	Node.js Native	27,899 req/s	36ms	Node.js
4	Rust (Actix)	27,458 req/s	36ms	Rust
5	Hono	23,325 req/s	43ms	Node.js
6	Express.js	8,361 req/s	119ms	Node.js
7	Next.js API	3,376 req/s	295ms	Node.js
8	Laravel Octane	1,124 req/s	877ms	PHP

Throughput Visualization

Bun Native      ████████████████████████████████████████████████  30,733 req/s
Go              ████████████████████████████████████████████      28,597 req/s
Node.js Native  ███████████████████████████████████████████       27,899 req/s
Rust (Actix)    ██████████████████████████████████████████        27,458 req/s
Hono            ████████████████████████████████████              23,325 req/s
Express.js      █████████████                                      8,361 req/s
Next.js API     █████                                              3,376 req/s
Laravel Octane  ██                                                 1,124 req/s

Test Methodology

Configuration

Parameter	Value
Virtual Users	1,000 concurrent users
Duration	30 seconds
Endpoint	/api/v1/health (simple JSON response)
Environment	Docker containers on local machine (Apple M1/M2)

What We Measured

• Throughput - Requests per second (req/s)
• Response Time - Average, p95, p99 latency
• Success Rate - Percentage of successful requests (HTTP 200)

Project Structure

script-k6-testing/
│
├── README.md
├── .gitignore
│
├── bun-test.js               # Bun native (port 3011)
├── go-test.js                # Go net/http (port 3006)
├── hono-test.js              # Hono (port 3010)
├── express-test.js           # Express.js (port 3005)
├── nextjs-test.js            # Next.js API (port 3009)
├── nodejs-native-test.js     # Node.js native (port 3008)
├── rust-actix-test.js        # Rust Actix-web (port 3007)
├── laravel-octane-test.js    # Laravel Octane (port 8000)
│
├── servers/
│   ├── bun/                  # Bun native server
│   ├── go/                   # Go server
│   ├── hono/                 # Hono framework server
│   ├── express/              # Express.js server
│   ├── nextjs/               # Next.js 14 with API routes
│   ├── nodejs-native/        # Node.js native HTTP server
│   └── rust-actix/           # Rust Actix-web server
│
└── _private/                 # Private test scripts (gitignored)

How to Run

Prerequisites

• Docker installed
• k6 installed

1. Start a Server

# Example: Start Bun server
cd servers/bun
docker-compose up -d --build

# Verify it's running
curl http://localhost:3011/api/v1/health

2. Run the Test

# Run k6 test (from root directory)
k6 run bun-test.js

3. Stop the Server

cd servers/bun
docker-compose down

Server Ports

Framework	Port
Express.js	3005
Go	3006
Rust (Actix)	3007
Node.js Native	3008
Next.js API	3009
Hono	3010
Bun Native	3011
Laravel Octane	8000

Key Insights

Why Bun is Fastest?

1. Zig-based runtime - Low-level language like Rust/C
2. JavaScriptCore engine - Safari's JS engine, optimized differently than V8
3. Built-in HTTP server - No external dependencies
4. Zero-copy optimizations - Minimal memory allocation

Node.js Framework Comparison

Framework	Throughput	vs Native
Node.js Native	27,899 req/s	1x
Hono	23,325 req/s	0.84x
Express.js	8,361 req/s	0.30x
Next.js API	3,376 req/s	0.12x

Performance Tiers

Tier	Throughput	Frameworks
Tier 1	25,000+ req/s	Bun, Go, Node.js Native, Rust
Tier 2	10,000 - 25,000 req/s	Hono
Tier 3	5,000 - 10,000 req/s	Express.js
Tier 4	< 5,000 req/s	Next.js API, Laravel Octane

Customizing Tests

Change VUs and Duration

export const options = {
  vus: 500,
  duration: '60s',
};

Stress Test with Stages

export const options = {
  stages: [
    { duration: '1m', target: 100 },
    { duration: '3m', target: 100 },
    { duration: '1m', target: 500 },
    { duration: '2m', target: 500 },
    { duration: '1m', target: 0 },
  ],
};

Disclaimer

⚠️ Hasil benchmark ini BUKAN tolak ukur mutlak untuk memilih framework.

Hasil testing ini hanya mengukur raw performance dari masing-masing framework dengan endpoint sederhana (health check). Performa aplikasi di dunia nyata dipengaruhi oleh banyak faktor lain:

Faktor	Pengaruh
Database	Query optimization, indexing, connection pooling
Caching	Redis, Memcached, in-memory cache
Server	CPU, RAM, SSD/NVMe, network bandwidth
Infrastructure	Load balancer, CDN, auto-scaling
Code Quality	Algorithm efficiency, N+1 queries, memory leaks
Business Logic	Complexity of operations, external API calls
Framework Features	ORM, validation, middleware, authentication

Catatan Penting

• Laravel mungkin lebih lambat di benchmark, tapi menyediakan fitur lengkap (ORM, queue, caching, authentication) yang mempercepat development
• Bun/Go/Rust sangat cepat, tapi butuh effort lebih untuk implementasi fitur yang sama
• Framework yang "lambat" di benchmark bisa sangat cepat di production dengan optimasi yang tepat
• Pilih framework berdasarkan kebutuhan project, skill tim, dan ekosistem, bukan hanya benchmark

Faktor yang Tidak Diuji

• Database queries (MySQL, PostgreSQL, MongoDB)
• File upload/download
• WebSocket connections
• Authentication & authorization
• Background jobs / queue processing
• Caching strategies
• Real network latency

License

MIT