Cloudflare Edge Ecosystem

⚡ Products — What Does What

Workers

Serverless functions at the edge

🛸

• Run JS/Wasm at Cloudflare's edge (300+ locations)
• Sub-ms cold starts, 128 MB memory
• 100k requests/day free

Best for: API endpoints, middleware, webhooks, auth, URL rewrites

Durable Objects

Strongly-consistent state + coordination

🧠

• Global singleton = one instance per namespace (or sharded)
• Real-time coordination: WebSocket sync, presence, locks
• Backed by a single-writer storage engine

Best for: collaborative state, game rooms, stock tickers, counters

Workers KV

Global, low-latency key-value store

🗄️

• Eventual consistency, cached at edge
• 1 GB storage, 1k writes/day, 100k reads/day free
• Values up to 25 MB, read in <10ms from cache

Best for: config, feature flags, redirects, static assets

R2

S3-compatible object storage

📦

• Zero egress fees — pay only for storage + operations
• 10 GB storage, 1M reads/month, 1M writes/month free
• Public buckets, presigned URLs, S3 API compatible

Best for: user uploads, backups, media files, log archives

D1

Serverless SQLite database

📊

• SQLite-compatible, read replicas at the edge
• 5 GB storage, 5M rows read/mo, 100k writes/mo free
• Migrations CLI, time-travel restore, HTTP API

Best for: relational data, user profiles, sessions, analytics

Queues

Message broker for async workloads

📨

• At-least-once delivery, batching, retries with backoff
• 1M operations/month free
• Consumers are Workers — your code handles the batch

Best for: background jobs, webhook delivery, decoupling services

Hyperdrive

Connection-pooled database proxy

🚀

• Connection pooling for Postgres/MySQL from Workers
• No open connections = no overhead between requests
• Caches query results at the edge

Best for: connecting Workers to an existing database

🏗️ Architecture — How It All Connects

graph TD User(("🌐 User")) subgraph CF[Cloudflare Edge] Workers[Workers
API / Middleware] KV[Workers KV
Global Cache] DO[Durable Objects
Coordinated State] Q[Queues
Message Broker] end subgraph Storage[Cloudflare Storage] R2[R2
Object Storage] D1[D1
SQLite Database] end subgraph External[External] PG[(Your Database)] Webhook[Webhook Endpoint] end User --> Workers Workers <--> KV Workers <--> DO Workers <--> D1 Workers <--> R2 Workers --> Q Q --> Workers Workers -- Hyperdrive --> PG Workers -- fetch() --> Webhook

Request path: User → Worker → (KV/D1/R2/DO) → Response

Async path: Worker → Queue → Worker → webhook/DB

External DB: Worker → Hyperdrive → Your database

🔁 Common Patterns

Upload → Transcode → Deliver

User uploads media, async pipeline processes it, result served from edge.

R2 (upload) → Worker → Queue → Worker → fetch() transcode server → R2 (output)

Real-time Collaboration

DO holds authoritative state, Workers bridge WebSocket connections.

DO (single-writer) ⟷ Workers (broadcast) ⟷ WebSocket clients

Static Site + API

Pages on the edge, Workers handle dynamic routes, D1 for data.

Cloudflare Pages (static) → Workers (API routes) → D1 (data)

Webhook Firehose → Buffer → Store

Queue buffers incoming webhooks, Workers batch-write to D1/R2.

Webhooks → Worker → Queue → Worker (batch) → D1

🚀 Homelab Project Ideas

📡

Server Health Dashboard

Workers ping your home lab endpoints every 5 minutes, push results to D1. Frontend reads from D1 via Workers. Alert on failure.

Workers D1 Cron Triggers

🗂️

Personal File Share

R2 bucket with a Worker that generates upload URLs. D1 tracks files. DO manages share links with expiry.

R2 D1 DO Workers

📰

News/Feed Aggregator

Cron Worker fetches RSS feeds, stores articles in D1. Workers API serves your personalized digest. R2 caches article images.

Workers D1 R2 Cron Triggers

🏓

WebSocket Game Lobby

DO manages game state. Workers handle WebSocket connections. D1 stores leaderboards. Queue processes deferred scoring.

DO Workers D1 Queues

📚 Matt Pocock — Cloudflare Learning Resources

🧑‍🏫 Matt's paid workshop

•

Making TypeScript Better for Cloudflare Workers on AI

Improving the DX for TypeScript developers writing Workers. Covers typed bindings, environment variables, and RPC-style DO calls.

•

TypeScript + D1: Type-Safe SQL

Using D1's type inference for end-to-end type safety from database to HTTP response. No ORM needed.

•

Building with Durable Objects & RPC

DO RPC makes calling Durable Objects feel like calling local functions. Matt's tutorial covers the DO lifecycle, alarms, and storage.

DO Tutorial → Workers Guide → D1 Type Safety →

🎁 Free Tier Limits at a Glance

Product	Free Tier	When to Upgrade
Workers	100k req/day	Deploy a busy API or cron-heavy app
KV	1 GB storage, 1k writes/day	Store user sessions or config at scale
R2	10 GB storage, 1M ops/month	Store videos or large file uploads
D1	5 GB, 5M rows read/mo	Production app with real traffic
Queues	1M operations/month	Heavy background job processing
DO	1M req/day, 5 GB storage	Real-time apps with millions of connections

📋 Quick Reference — When to Use What

I need to...

→ Run serverless functions → Workers

→ Store user sessions → KV

→ Store files/users upload → R2

→ Query relational data → D1

→ Coordinate real-time state → DO

→ Decouple async work → Queues

→ Connect to my existing DB → Hyperdrive

→ Build a full app → Workers + D1 + R2