bench

SSR framework benchmark comparing One, TanStack Start, Next.js, and React Router.

Based on Platformatic's benchmark methodology, extended with a One implementation and local benchmarking.

Application

Each framework implements the same e-commerce trading card marketplace with:

• 5 games, 50 sets, 10,000 cards, 100 sellers, 100,000+ listings
• 9 SSR routes (homepage, search, card detail, game detail, games list, sellers, set detail, cart, seller detail)
• JSON file-based database with configurable simulated latency (1-5ms)
• Identical data, identical UI, identical server-side work

Frameworks

Directory	Framework	Server
one/	One	Hono
tanstack/	TanStack Start	Nitro/h3
next/	Next.js	next start
react-router/	React Router v7	Express

Quick Start

Local Benchmark (One vs TanStack)

# install dependencies
cd one && npm install && cd ..
cd tanstack && npm install && cd ..

# build both
cd one && npm run build && cd ..
cd tanstack && npm run build && cd ..

# run benchmark
./bench-local.sh

Configuration

Variable	Default	Description
DURATION	30	Seconds per route test
CONNECTIONS	100	Concurrent connections
DB_DELAY_ENABLED	true	Enable simulated DB latency
DB_DELAY_MIN	1	Minimum delay (ms)
DB_DELAY_MAX	5	Maximum delay (ms)

EKS Benchmark (all frameworks)

For cloud benchmarking on AWS EKS with k6 load testing, see the EKS benchmark docs and infrastructure setup.

AWS_PROFILE=<profile> FRAMEWORK=next ./benchmark.sh

Data

Shared test data lives in data/ and is symlinked into each framework directory. Generate fresh data with:

npx tsx lib/generate-data.ts

Results

See BENCHMARK-TABLES.md for benchmark results and feature comparison.

License

MIT