惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Martin Fowler
Martin Fowler
大猫的无限游戏
大猫的无限游戏
J
Java Code Geeks
罗磊的独立博客
雷峰网
雷峰网
G
Google Developers Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
爱范儿
爱范儿
B
Blog RSS Feed
腾讯CDC
Apple Machine Learning Research
Apple Machine Learning Research
D
Docker
Recent Announcements
Recent Announcements
T
Tailwind CSS Blog
博客园 - 聂微东
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
Vercel News
Vercel News
小众软件
小众软件
人人都是产品经理
人人都是产品经理
云风的 BLOG
云风的 BLOG
IT之家
IT之家
Blog — PlanetScale
Blog — PlanetScale
I
InfoQ
S
SegmentFault 最新的问题

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
Standing on shoulders: the stack that makes Floci start i...
Hector Ventu · 2026-05-22 · via DEV Community

Floci starts in 24ms and idles at 13 MiB of RAM.

That number gets attention, and it tends to obscure the real story. It isn't about one clever optimization. It's about four open-source projects that have been quietly perfecting their respective jobs for years, stacked on top of each other in a way that makes the whole thing possible.

Floci is a free, MIT-licensed AWS emulator. It speaks the AWS wire protocols on localhost:4566, emulates 45 services, and is verified by 1,925 SDK compatibility tests across Java, Python, Node.js, Go, and Rust. Here's the stack underneath it.

Netty, the async I/O substrate

Netty has been around since 2004. If you've used Elasticsearch, Cassandra, gRPC-Java, Pulsar, Spark, the AWS SDK v2 async client, or Spring WebFlux (via Reactor Netty), you've used Netty. Every TCP connection into Floci, every UDP datagram, every connection to a spawned RDS container terminates in a Netty channel pipeline somewhere.

Two decades of hardening by Apple, LinkedIn, Twitter, and basically every JVM-based hyperscaler. The bugs at the protocol-decoder level have been found and fixed. The performance characteristics are extremely well understood. And critically, Netty has been a good citizen of the native-image world. It plays well with GraalVM out of the box.

Vert.x, the reactive toolkit

Vert.x is the layer where Floci uses things directly. It's Netty with ergonomics: an event loop, a router, protocol-specific APIs for HTTP, DNS, TCP, WebSockets, gRPC, all sharing the same threading model.

Floci uses Vert.x for two things that matter.

The HTTP router. AWS routing is weird. The same TCP endpoint serves REST-XML (S3), REST-JSON, x-amz-json-1.0 and 1.1 (DynamoDB), query-string protocol (older services), and EC2 query protocol. Vert.x's router gives you full control over dispatch without forcing a single framework's idea of what a controller looks like.

The embedded DNS server. Floci runs its own authoritative DNS on UDP/53 inside the container, so *.localhost.floci.io resolves to Floci's container IP on the Docker network. That's how virtual-hosted S3 (https://my-bucket.s3.localhost.floci.io/key) works from inside spawned Lambda containers with no extra setup. Tractable in Vert.x because HTTP, UDP, and TCP all share infrastructure: same event loop, same lifecycle. About 200 lines.

Quarkus, the framework layer

This is where Floci becomes a coherent application instead of a pile of libraries.

Quarkus does at compile time what traditional frameworks do at runtime: bean discovery, dependency injection wiring, configuration resolution, JAX-RS route registration, Jackson serializer generation. That matters because GraalVM is a closed-world compiler. It has to know at compile time every class that might be loaded and every method that might be called via reflection. Quarkus is engineered from the ground up to give the native compiler that information.

The validating case for Quarkus is Keycloak, which migrated its entire codebase from WildFly to Quarkus in 2023, explicitly citing startup time and memory. When a project of that scale and conservatism makes that move, it's a strong signal.

GraalVM Mandrel, the native-image compiler

This is the layer that delivers the headline numbers. Ahead-of-time compilation. Closed-world analysis. Dead code elimination. The output is a single statically-linked native executable. No JVM at runtime, no JIT warmup, no classpath scanning at startup.

Metric Floci (native) Typical JVM
Startup ~24 ms 2,000 to 6,000 ms
Idle RAM ~13 MiB 200 to 400 MiB

Mandrel is Red Hat's downstream of GraalVM, tuned for Quarkus workloads, with release cadence aligned to Quarkus. An underrated piece of supply-chain hygiene.

Thank you

Floci wouldn't exist without any of these layers. Forty-five AWS services is a lot of surface area for one maintainer. The reason it's tractable is that the four layers below don't need maintenance. They're production-hardened by people I'll never meet, and I get to focus on AWS wire-protocol fidelity, which is the interesting part of an emulator.

Thank you to the Netty, Vert.x, Quarkus, and GraalVM teams. Floci stands on your shoulders.

github.com/floci-io/floci