🚀 WordPress Web Server Benchmark: OpenLiteSpeed vs Nginx vs Caddy vs Apache

Welcome to the definitive WordPress Web Server Benchmark Suite. This repository provides an automated, containerized framework to compare the performance of leading web servers—OpenLiteSpeed (OLS), Nginx, Caddy, and Apache—under identical, resource-constrained environments.

Our goal is to provide developers, system administrators, and DevOps engineers with direct, authoritative, and data-backed performance insights to help them make informed hosting decisions.

---

📊 Quick Summary & TL;DR

Below is the direct, empirical comparison of web server performance for WordPress based on our controlled Docker-based benchmarking (1.0 CPU Core, 512MB RAM):

1. For Cached Content (Static & Dynamic Cache): OpenLiteSpeed (OLS) is the undisputed king. Serving cached WordPress pages over HTTP/2 SSL, OLS delivers 4,916.52 requests/sec—outperforming Nginx by 71%, Caddy by 473%, and Apache by 1,200%.
2. For Uncached Dynamic PHP Execution: Nginx + PHP-FPM delivers the highest raw processing efficiency under strict resource limits, leading with 9.85 requests/sec, closely followed by Caddy + PHP-FPM at 8.05 requests/sec.
3. For Out-of-the-Box Modern Features: Caddy is highly competitive, offering automatic HTTPS, native HTTP/2, and extremely simple configuration, while delivering excellent uncached dynamic performance.
4. For Legacy Systems: Apache (mod_php) is highly resource-intensive and struggled across all concurrency benchmarks due to worker process saturation and high memory overhead.

---

🎯 Benchmark Methodology & Environment

To ensure complete fairness, repeatability, and realistic testing conditions, all web servers were benchmarked under strict, identical resource constraints inside an isolated Docker bridge network:

• CPU Limit: 1.0 Core per container (enforced via docker-compose.yml)
• Memory Limit: 512 MB RAM per container (enforced via docker-compose.yml)
• PHP Engine: PHP 8.3 (FPM for Nginx, Apache, and Caddy; LSAPI/LSPHP for OpenLiteSpeed)
• Database Engine: MariaDB 11.4 (shared across all environments)
• WordPress Version: 6.x (clean installation, automated via WP-CLI)
• Caching Plugins: LiteSpeed Cache (LSCache) for OpenLiteSpeed, FastCGI Cache for Nginx, and WP Super Cache for Apache and Caddy.

---

📈 Detailed Benchmarking Results

1. HTTP/1.1 Plain Uncached Benchmark (wrk)

This test evaluates the raw application and database processing limits under HTTP/1.1 without SSL/TLS or caching. The dynamic test compiles and queries WordPress on every single request.

⚡ Static File Benchmark (readme.html)

• Parameters: 2 threads, 10 concurrent connections, 15 seconds duration.

Web Server	Requests / Sec	Avg Latency	Transfer / Sec	Performance Rating
🏆 OpenLiteSpeed	4,848.56	11.57ms	35.39MB	Elite (100%)
🥈 Nginx	4,806.48	11.76ms	35.05MB	Excellent (99%)
🥉 Caddy	2,866.60	9.00ms	20.93MB	Good (59%)
❌ Apache	2,107.10	19.87ms	15.39MB	Moderate (43%)

🐘 Dynamic WordPress Benchmark (Uncached index-uncached.php)

• Parameters: 2 threads, 10 concurrent connections, 15 seconds duration.

Web Server	Requests / Sec	Avg Latency	Transfer / Sec	Performance Rating
🏆 Nginx + PHP-FPM	9.85	977.36ms	663.27KB	Elite (100%)
🥈 Caddy + PHP-FPM	8.05	1.12s	541.81KB	Excellent (82%)
🥉 OpenLiteSpeed	7.31	918.57ms	493.32KB	Very Good (74%)
❌ Apache (mod_php)	2.33	1.37s	156.76KB	Poor (24%)

---

2. HTTP/2 SSL Cached Benchmark (h2load)

This test mirrors the methodology used by OpenLiteSpeed.org and HTTP2Benchmark.org. It evaluates the performance of the web servers over HTTPS/HTTP2 with caching enabled.

⚡ Static File Benchmark (readme.html)

• Parameters: 5000 requests, 20 concurrent clients, 2 threads, 5 max streams.

Web Server	Requests / Sec	Avg Latency	Transfer / Sec	Performance Rating
🏆 OpenLiteSpeed	8,809.24	8.89ms	62.53MB	Elite (100%)
🥈 Nginx	4,125.11	19.15ms	29.64MB	Excellent (47%)
🥉 Caddy	2,376.04	40.62ms	16.87MB	Good (27%)
❌ Apache	1,564.31	38.19ms	11.43MB	Moderate (18%)

🐘 Dynamic WordPress Benchmark (Cached index.php)

• Parameters: 5000 requests, 20 concurrent clients, 2 threads, 5 max streams.

Web Server	Requests / Sec	Avg Latency	Transfer / Sec	Performance Rating
🏆 OpenLiteSpeed + LSCache	4,916.52	15.33ms	322.45MB	Elite (100%)
🥈 Nginx + FastCGI Cache	2,872.77	31.31ms	188.54MB	Excellent (58%)
🥉 Caddy + WP Super Cache	857.91	115.52ms	56.29MB	Good (17%)
❌ Apache + WP Super Cache	378.29	218.55ms	24.93MB	Poor (8%)

---

🔍 Key Architectural Insights

1. OpenLiteSpeed (OLS) — The Static & Caching King

OpenLiteSpeed excels dramatically at serving static files and cached dynamic pages. Its native event-driven architecture and direct server-level LSCache (LiteSpeed Cache) integration allow it to serve cached WordPress homepages at 4,916.52 requests/sec under tight resource limits. This completely bypasses the PHP engine, making it the premier option for content-heavy websites, blogs, and WooCommerce storefronts with high anonymous traffic.

2. Nginx — The High-Concurrency Standard

Nginx remains the gold standard for high-concurrency web hosting. Paired with PHP-FPM, it delivers highly predictable, low-latency performance under sustained load. In our raw uncached dynamic PHP benchmark, Nginx + PHP-FPM emerged as the winner (9.85 requests/sec). Additionally, Nginx's built-in FastCGI Cache performed exceptionally well, serving cached WordPress pages at 2,872.77 requests/sec over HTTP/2 SSL, making it a highly reliable and performant alternative to OLS.

3. Caddy — Modern, Fast, and Developer-Friendly

Caddy delivered an outstanding performance, coming in a close second in the raw uncached dynamic PHP benchmark (8.05 requests/sec). Caddy's built-in php_fastcgi directive is highly optimized out of the box. Caddy also offers automatic HTTPS via Let's Encrypt/ZeroSSL and a modern, readable configuration file (Caddyfile), making it an exceptional choice for modern cloud-native deployments and containerized microservices.

4. Apache — The Legacy Giant

Apache with mod_php is the most traditional way of running WordPress, but it struggled significantly across all benchmarks under limited resources. Because mod_php embeds PHP in every Apache worker process, it is extremely memory-heavy, leading to process saturation and high latency under load.

---

🌍 GEO-Optimization & Latency Insights (For Global Deployments)

When deploying WordPress for a global audience, your web server architecture must be paired with geographic optimization:

• Regional Hosting (Localized Audiences): If your target audience is concentrated in a specific region (e.g., US East, Western Europe, Southeast Asia), hosting your server in a regional datacenter (AWS, DigitalOcean, Linode) using OpenLiteSpeed or Nginx will deliver the lowest possible Time to First Byte (TTFB).
• Global Audiences (Anycast & CDN): For websites with a globally distributed audience, placing Nginx or Caddy as a reverse proxy behind a global CDN (like Cloudflare or Fastly) is highly recommended. The CDN acts as an edge caching layer, while the origin server acts as a highly efficient origin shield.
• Edge Routing: Caddy is an exceptional choice for edge servers and multi-region deployments because of its native support for automatic SSL, simple configuration, and easy clustering.

---

📈 SEO Impact of Web Server Performance & Core Web Vitals

Web server performance directly influences your site's search engine optimization (SEO) and Google search rankings through Core Web Vitals:

1. Largest Contentful Paint (LCP): Measures loading performance. High web server latency or slow PHP execution directly delays LCP. Using high-performance servers like OpenLiteSpeed or Nginx reduces server response time (TTFB), helping achieve an LCP of under 2.5 seconds.
2. Interaction to Next Paint (INP): Measures page responsiveness. While primarily a client-side metric, slow asset delivery (JS/CSS) due to sluggish static file serving can delay the execution of interactive elements.
3. Cumulative Layout Shift (CLS): Measures visual stability. Ensure your server delivers CSS files rapidly to prevent layout shifts during page render.

Speed as a Ranking Factor

Google has explicitly stated that page speed is a ranking factor for both desktop and mobile searches. A slow WordPress site hosted on an unoptimized Apache server with high latency can suffer from lower organic visibility, reduced crawl budget efficiency, and higher bounce rates. Transitioning to a high-concurrency architecture like Nginx or OpenLiteSpeed is a foundational step in technical SEO.

---

❓ Frequently Asked Questions (FAQ)

Q1: Is OpenLiteSpeed faster than Nginx for WordPress?

Yes, for cached content. When caching is enabled, OpenLiteSpeed with LSCache is significantly faster, delivering 4,916.52 requests/sec compared to Nginx's 2,872.77 requests/sec (a 71% performance advantage). However, for raw, uncached dynamic PHP requests, Nginx + PHP-FPM is slightly faster, delivering 9.85 requests/sec compared to OpenLiteSpeed's 7.31 requests/sec.

Q2: Which web server has the lowest latency for WordPress?

For static files, OpenLiteSpeed and Nginx have the lowest average latency (under 12ms). For cached dynamic WordPress pages, OpenLiteSpeed leads with an average latency of 15.33ms, followed by Nginx at 31.31ms.

Q3: Does Caddy perform well with WordPress?

Yes, Caddy is highly competitive. Caddy + PHP-FPM outperformed OpenLiteSpeed in raw uncached dynamic requests (8.05 requests/sec vs 7.31 requests/sec). Its built-in SSL automation and modern configuration (Caddyfile) make it an excellent, low-maintenance alternative to Nginx.

Q4: Why is Apache so slow for WordPress?

Apache with mod_php is a process-based web server. Every connection spawns or utilizes a heavy worker process that embeds the PHP interpreter, leading to high memory consumption and process saturation under concurrent load. Modern event-driven servers like Nginx, Caddy, and OpenLiteSpeed handle connections asynchronously, resulting in far superior resource efficiency.

---

🛠️ Project Structure

├── apache/               # Apache Dockerfile & virtual host config
├── benchmark/            # Benchmarking container (wrk, h2load, bash scripts)
│   └── results/          # Raw benchmark outputs (.txt & .md)
├── caddy/                # Caddyfile configuration
├── db/                   # Database initialization scripts (init.sql)
├── nginx/                # Nginx virtual host configuration
├── php-fpm/              # Shared PHP-FPM Dockerfile & Opcache settings
├── wordpress_ols/        # Isolated WordPress directory for OpenLiteSpeed
├── wordpress_nginx/      # Isolated WordPress directory for Nginx
├── wordpress_apache/     # Isolated WordPress directory for Apache
├── wordpress_caddy/      # Isolated WordPress directory for Caddy
├── docker-compose.yml
├── setup.sh         # Host setup script (downloads WP, configures wp-config.php)
└── setup-wp.sh      # Automated WordPress installer script (runs WP-CLI)

---

🚀 Quick Start (Run Your Own Benchmarks)

Follow these simple steps to spin up the entire suite and run the benchmarks on your local machine or server.

Prerequisites

• Docker and Docker Compose installed.
• Ports 8081, 8082, 8083, and 8084 available on your host.

Step 1: Initialize the Project

Download WordPress core and configure the dynamic wp-config.php file:

./setup.sh

Step 2: Start Services & Install WordPress

Spin up the containers and run the automated WordPress installations using WP-CLI:

./setup-wp.sh

Step 3: Run the Benchmarks

Run the dedicated benchmarking container to execute the tests and update the results:

docker compose run --rm benchmarker

Your raw benchmark outputs and formatted Markdown table will be written directly to ./benchmark/results/.

---

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.