Server infrastructure

rari benchmark

React Server Components on a Rust runtime. No Node.js in the request hot path. Measuring real-world performance on a $0 EC2 t3.micro.

Live on virginia.uptimeclaw.com

This app runs on a $0 AWS t3.micro (1 vCPU, 1GB RAM) with the rari Rust runtime. No Node.js in the request hot path — the Rust binary handles all incoming requests.

~137 MB

Runtime RSS

0.7%

CPU at idle

~476ms

Build time

Measured April 2026 on EC2 t3.micro us-east-1

Response Time (Single Request)

MetricrariNext.jsImprovement
Avg Response0.08ms0.78ms9.8x faster
P95 Latency0.12ms0.86ms7.2x faster
Bundle Size281 KB634 KB56% smaller

Throughput Under Load (50 connections, 30s)

MetricrariNext.jsImprovement
Requests/sec71,4041,49747.7x higher
Avg Latency0.70ms33.42ms47.7x faster
P95 Latency1.19ms41.90ms35.2x faster

Why rari?

Rust Runtime

Rust Runtime

HTTP server, RSC renderer, and routing written in Rust with embedded V8. No Node.js in the request hot path.

React Server Components
🧩

React Server Components

Server components by default. Use 'use client' only when you need interactivity. True RSC semantics.

Zero Cold Start
🔥

Zero Cold Start

Persistent process — always warm, always fast. No serverless spin-up penalty.

🌊

Streaming SSR

Progressive rendering with Suspense boundaries. Users see content immediately.

📦

Tiny Bundles

56% smaller client JavaScript than Next.js. Rolldown-powered Vite builds in under 2 seconds.

🏠

Self-Hosted

Run on your own hardware. No vendor lock-in, no cold starts, no per-request billing.

Test the API

Hit these endpoints directly to see raw response times:

Server-rendered at 2026-05-26T12:20:14.447Z