wrk - HTTP Benchmarking Tool

Prerequisites

• Operating System: Linux, macOS, or other Unix-like systems
• Compiler: GCC or Clang (C compiler required)
• Build System: Make
• Optional: LuaJIT (for scripting support)
• System Configuration: Sufficient ephemeral ports and proper socket recycling settings

Installation

Building from Source

1. Clone the repository:

   git clone https://github.com/wg/wrk.git
   cd wrk
   

2. Build the project:

   make
   

3. (Optional) Install to system:

   sudo make install
   

System Configuration

For optimal performance, ensure your system has sufficient ephemeral ports:

# Check current ephemeral port range
sysctl net.ipv4.ip_local_port_range

# Increase if necessary
sudo sysctl -w net.ipv4.ip_local_port_range="1024 65535"

Configuration

Command Line Options

• -t, --threads: Number of threads to use (default: 1)
• -c, --connections: Total HTTP connections to keep open (default: 10)
• -d, --duration: Test duration (e.g., 30s, 2m, 2h)
• -s, --script: LuaJIT script for custom request generation
• -H, --header: Custom HTTP headers to add to requests
• --latency: Print detailed latency statistics
• --timeout: Response timeout threshold

Environment Variables

No specific environment variables are required for basic operation.

Lua Scripting

For advanced usage with Lua scripts:

1. Ensure LuaJIT is installed
2. Place custom scripts in the scripts/ directory or specify with -s flag

Build & Run

Development

1. Build the project:

   make
   

2. Run a basic benchmark:

   ./wrk -t12 -c400 -d30s http://127.0.0.1:8080/index.html
   

Production

For production benchmarking:

1. Ensure system is properly configured (see Prerequisites)
2. Run with appropriate thread and connection counts:

   wrk -t$(nproc) -c1000 -d5m --latency http://your-server.com/
   

Server Configuration

For the server being tested:

• Set listen backlog greater than concurrent connections
• Ensure proper socket recycling
• Monitor system resources during testing

Deployment

wrk is a command-line tool that doesn't require traditional deployment. For distributed testing:

Local Deployment

1. Build and install on each testing machine
2. Coordinate tests from a central location
3. Aggregate results from multiple nodes

Cloud Deployment

For cloud-based testing:

1. Deploy to multiple cloud instances
2. Use SSH or orchestration tools to coordinate tests
3. Consider using containerization for consistent environments

Troubleshooting

Common Issues

1. "Address already in use" errors

• Solution: Increase ephemeral port range or wait for sockets to timeout

2. Insufficient connections

• Solution: Check system limits and increase net.ipv4.ip_local_port_range

3. Poor performance

• Check: CPU affinity, network interface saturation, system resource limits

4. Lua script errors

• Verify: LuaJIT installation and script syntax

Performance Tuning

1. Thread Count: Match to CPU cores
2. Connection Count: Balance with system resources
3. Duration: Long enough for steady state (typically 30+ seconds)

Monitoring

During tests, monitor:

• CPU usage
• Network throughput
• System load
• Memory usage

Validation

Verify results by:

1. Running multiple tests
2. Checking for consistent metrics
3. Comparing with other benchmarking tools if needed