Benchmarks

Performance is the cornerstone of Dapil. These benchmarks compare Dapil against industry standards and other high-performance frameworks.

Hardware & Environment

• CPU: x86_64 Linux
• Concurrency: 50 concurrent connections
• Total Requests: 10,000
• Build Mode: Release (LTO enabled, 1 codegen unit)

Throughput Comparison

Framework	Requests/Sec	Latency (Mean)
Dapil (Extreme)	29,661	1.6 ms
Django-Bolt	19,511	2.5 ms
Django (Gunicorn)	5,623	8.8 ms
FastAPI (Uvicorn)	3,563	14.0 ms

Analysis

1. Vs FastAPI (8.32x): FastAPI (and Uvicorn) are bottlenecked by Python's asynchronous overhead and GIL management when handling concurrent I/O. Dapil offloads all I/O to Rust, leaving Python only responsible for the business logic.
2. Vs Django-Bolt (1.52x): While Django-Bolt also uses a Rust core, Dapil's specialized Single Actor worker model provides lower overhead for simple dispatching, ensuring the GIL is held optimally.
3. Vs Standard Django (5.27x): Standard sync Django is limited by worker process/thread overhead. Dapil's hybrid model provides the efficiency of an async event loop with the simplicity of sync handlers.

Reproducing Benchmarks

We use ab (Apache Benchmark) for reliable throughput measurement.

1. Run the Dapil server

python examples/hello_world/main.py

2. Execute Benchmark

ab -n 10000 -c 50 http://127.0.0.1:8181/

3. Comparison Scripts

You can find the scripts used for other frameworks in the benchmarks/ directory: - fastapi_app.py - django_app.py