Benchmarks

Benchmark numbers are useful only when they are reproducible and tied to a specific environment.

Latest sample baseline (February 19, 2026)

Command:

hatch run python benchmarks/run.py --http-requests 5000

Sample output:

Scenario	Server	Operations	Duration (s)	Ops/s
http	palfrey	5000	0.1426	35063.12
http	uvicorn	5000	0.2721	18374.60
websocket	palfrey	1000	0.0306	32631.40
websocket	uvicorn	1000	0.0702	14235.33

Relative throughput in this run:

• http: 1.908x (Palfrey / Uvicorn)
• websocket: 2.292x (Palfrey / Uvicorn)

Important: These numbers are environment-specific and not universal guarantees.

Benchmark principles

• compare equivalent runtime modes
• keep commands and environment details explicit
• run multiple samples and inspect variance
• include failure/error counts, not only throughput

Suggested scenario matrix

1. small JSON API, high concurrency
2. mixed payload sizes and keep-alive reuse
3. websocket message throughput
4. reverse-proxy deployment path

Reporting template

For each scenario, record:

• command line
• hardware and OS
• Python and dependency versions
• operations/sec
• p50/p95/p99 latency
• CPU and memory
• error count

Communication rule

Never claim a performance improvement without the reproducible command, environment details, and raw results.