DEV Community

Aleix Nguyen
Aleix Nguyen

Posted on

🚀 5,000 Concurrent WordPress Users on a $6 VPS – Docker Stack Case Study

#docker #wordpress #nginx #performance

“5000 users. 187ms response. 431MB RAM. 2–6% CPU. $6/month. We didn’t break the laws of physics — we just understood them better.”

🎯 Goal

Can WordPress serve 5000 concurrent users on a 1GB RAM VPS... and stay under 200ms?

  • VPS: Vultr (1vCPU, 1GB RAM – $6/mo)
  • Stack: 100% open-source, containerized
  • Caching: 4 layers (CDN, full-page, object, opcode)
  • Budget: Under $10/mo total

✅ Spoiler: Yes. Wildly yes.

⚡ The Benchmark

Users Response RAM CPU
250 198ms 406MB ~2%
500 191ms 399MB ~2%
750 191ms 403MB ~2–5%
1500 192ms 416MB ~3-6%
5000 187ms (max: 391ms) 431MB 2–6% ✅

2600% traffic ⬆️ = minimal latency ⬆️

Sub-linear scaling = confirmed

🧱 Stack Overview

  • NGINX (Alpine)
  • PHP 8.2-FPM + OPCache
  • WordPress (latest)
  • MariaDB (Docker image)
  • Redis (object cache)
  • Caddy (SSL + HTTP/3)
  • Docker Compose
  • Cloudflare (free tier)

🔁 Cache Hierarchy

Layer Hit Rate Notes
Cloudflare 45.8% CDN edge cache
Cache Enabler 87.2% Full-page HTML
Redis 99.93% Object, transients
OPcache 96.7% PHP bytecode

➡️ Only ~0.4% hit PHP, and ~0.02% hit the database.

🧠 PHP-FPM Config (for 1GB RAM) 

pm = dynamic
pm.max_children = 4
pm.start_servers = 2
pm.min_spare_servers = 1
pm.max_spare_servers = 3
pm.max_requests = 500
Enter fullscreen mode Exit fullscreen mode

🔥 OPCache 

opcache.enable=1
opcache.memory_consumption=128
opcache.max_accelerated_files=10000
opcache.validate_timestamps=0
Enter fullscreen mode Exit fullscreen mode

⚙️ Redis 

maxmemory 256mb
maxmemory-policy allkeys-lru
Enter fullscreen mode Exit fullscreen mode

🌐 NGINX + Caddy

  • Gzip enabled
  • Static cache: 30d
  • HTML cache: 1h
  • HTTP/3 via Caddy + TLS

📊 Performance Breakdown

Layer Served Hit Rate
Cloudflare 2,290 users 45.8%
Cache Enabler 2,070 users 87.2%
PHP+Redis 639 users 99.93%
DB (MariaDB) 1 query! 🤯

Memory Allocation

  • PHP-FPM + WP: ~180MB
  • Redis: ~128MB
  • MariaDB: ~80MB
  • NGINX + Caddy: ~27MB
  • System overhead: ~15MB

🧠 Why It Works

✅ 4-layer caching synergy

✅ Tuned PHP-FPM for RAM budget

✅ Zero waste CPU cycles

✅ Sub-linear scaling due to warm cache

✅ $6 infra instead of $200+ managed hosting

📦 Repo & Setup

🔗 GitHub: wordpress-docker-performance

  1. Clone Docker Compose setup
  2. Launch with docker-compose up -d
  3. Install:
    • Cache Enabler
    • Redis Object Cache
  4. Configure wp-config.php for Redis
  5. Done. Cache hit >99% 🎉

💬 What do you think?

  • Would you run WP this way in prod?
  • Can we push this to 8000+ concurrent users?

👇 Drop your thoughts in the comments — let’s scale smarter!

Top comments (0)