Why I Chose Cloudflare Workers Over Traditional Servers

For the first year of SDBOTS, I ran everything on a €10/month VPS. It worked. Then we started seeing 50–200ms response times from users in Southeast Asia and the US, and I started looking at alternatives.

The Migration Numbers

After moving the API endpoints to Cloudflare Workers, median response time dropped from ~120ms to ~18ms globally. The 99th percentile improved even more — from 800ms to ~60ms.

• P50: 120ms → 18ms
• P99: 800ms → 60ms
• Monthly cost: €10 → $0 (free tier)
• Uptime: 99.2% → 99.99% (Cloudflare SLA)

What I Gave Up

Workers have a 128MB memory limit, no filesystem, no long-running processes, and a 50ms CPU time limit on the free plan. If you need to do heavy computation — image processing, ML inference — you still need a traditional server or a queue.

My Hybrid Setup

I use Workers for all stateless API routes (auth, lookup, bot commands) and keep a small VPS for background jobs — sending scheduled messages, running ML pipelines, processing large file uploads. Best of both worlds.